GS/DX11: Align drawlist bounding box to 4 pixel boundaries.

Will be faster using Direct Memory Access, otherwise it may stall as more commands need to be issued.

bin/resources/GameIndex.yaml

@@ -24781,6 +24781,11 @@ SLES-53820:
 SLES-53821:
   name: "Radio Helicopter II"
   region: "PAL-E"
+  patches:
+    9A695202:
+      content: |-
+        comment=Patch that nops a branch instruction causing a freeze.
+        patch=1,EE,001799AC,word,00000000
 SLES-53824:
   name: "Trapt"
   region: "PAL-E"
@@ -40089,6 +40094,11 @@ SLPM-62624:
   name-sort: "ぷちこぷたー2"
   name-en: "Petit Copter 2"
   region: "NTSC-J"
+  patches:
+    9A695202:
+      content: |-
+        comment=Patch that nops a branch instruction causing a freeze.
+        patch=1,EE,001799B8,word,00000000
 SLPM-62625:
   name: "鬼浜爆走愚連隊 激闘編"
   name-sort: "おにはまばくそうぐれんたい げきとうへん"

pcsx2/GS/Renderers/DX11/GSDevice11.cpp

@@ -2847,6 +2847,7 @@ void GSDevice11::SendHWDraw(const GSHWDrawConfig& config, GSTexture* draw_rt_clo
 		{
 			const u32 draw_list_size = static_cast<u32>(config.drawlist->size());
 			const u32 indices_per_prim = config.indices_per_prim;
+			const GSVector4i rtsize(0, 0, draw_rt->GetWidth(), draw_rt->GetHeight());
 
 			pxAssert(config.drawlist && !config.drawlist->empty());
 			pxAssert(config.drawlist_bbox && static_cast<u32>(config.drawlist_bbox->size()) == draw_list_size);
@@ -2855,10 +2856,28 @@ void GSDevice11::SendHWDraw(const GSHWDrawConfig& config, GSTexture* draw_rt_clo
 			{
 				const u32 count = (*config.drawlist)[n] * indices_per_prim;
 
-				GSVector4i bbox = (*config.drawlist_bbox)[n].rintersect(config.drawarea);
+				const GSVector4i original_bbox = (*config.drawlist_bbox)[n].rintersect(config.drawarea);
+				GSVector4i snapped_bbox = original_bbox;
 
-				// Copy only the part needed by the draw.
-				CopyAndBind(bbox);
+				// We don't want the snapped box adjustments when the rect is empty as it might make the copy to pass.
+				// The empty rect itself needs to be handled in renderer properly.
+				if (!snapped_bbox.rempty())
+				{
+					// Aligning bbox to 4 pixel boundaries so copies will be faster using Direct Memory Access,
+					// otherwise it may stall as more commands need to be issued.
+					snapped_bbox.left &= ~3;
+					snapped_bbox.top &= ~3;
+					snapped_bbox.right = (snapped_bbox.right + 3) & ~3;
+					snapped_bbox.bottom = (snapped_bbox.bottom + 3) & ~3;
+
+					// Ensure the new sizes are within bounds.
+					snapped_bbox.left = std::max(0, snapped_bbox.left);
+					snapped_bbox.top = std::max(0, snapped_bbox.top);
+					snapped_bbox.right = std::min(snapped_bbox.right, rtsize.right);
+					snapped_bbox.bottom = std::min(snapped_bbox.bottom, rtsize.bottom);
+				}
+
+				CopyAndBind(snapped_bbox);
 
 				DrawIndexedPrimitive(p, count);
 				p += count;

pcsx2/GS/Renderers/HW/GSRendererHW.cpp

@@ -5910,6 +5910,7 @@ void GSRendererHW::EmulateBlending(int rt_alpha_min, int rt_alpha_max, const boo
 	{
 		case AccBlendLevel::Maximum:
 			sw_blending |= true;
+			accumulation_blend &= (GSLocalMemory::m_psm[m_cached_ctx.FRAME.PSM].bpp == 32);
 			[[fallthrough]];
 		case AccBlendLevel::Full:
 			sw_blending |= m_conf.ps.blend_a != m_conf.ps.blend_b && alpha_c0_high_max_one;

pcsx2/SPU2/Dma.cpp

@@ -163,7 +163,7 @@ void V_Core::StartADMAWrite(u16* pMem, u32 sz)
 	if ((AutoDMACtrl & (Index + 1)) == 0)
 	{
 		ActiveTSA = 0x2000 + (Index << 10);
-		DMAICounter = size * 4;
+		DMAICounter = size * 48;
 		LastClock = psxRegs.cycle;
 	}
 	else if (size >= 256)
@@ -191,7 +191,7 @@ void V_Core::StartADMAWrite(u16* pMem, u32 sz)
 		if (SPU2::MsgToConsole())
 			SPU2::ConLog("ADMA%c Error Size of %x too small\n", GetDmaIndexChar(), size);
 		InputDataLeft = 0;
-		DMAICounter = size * 4;
+		DMAICounter = size * 48;
 		LastClock = psxRegs.cycle;
 	}
 }
@@ -248,7 +248,7 @@ void V_Core::FinishDMAwrite()
 		DMA7LogWrite(DMAPtr, ReadSize << 1);
 #endif
 
-	u32 buff1end = ActiveTSA + std::min(ReadSize, (u32)0x100 + std::abs(DMAICounter / 4));
+	u32 buff1end = ActiveTSA + std::min(ReadSize, (u32)0x100 + std::abs(DMAICounter / 48));
 	u32 buff2end = 0;
 	if (buff1end > 0x100000)
 	{
@@ -343,7 +343,7 @@ void V_Core::FinishDMAwrite()
 	DMAPtr += TDA - ActiveTSA;
 	ReadSize -= TDA - ActiveTSA;
 
-	DMAICounter = (DMAICounter - ReadSize) * 4;
+	DMAICounter = (DMAICounter - ReadSize) * 48;
 
 	CounterUpdate(DMAICounter);
 
@@ -354,7 +354,7 @@ void V_Core::FinishDMAwrite()
 
 void V_Core::FinishDMAread()
 {
-	u32 buff1end = ActiveTSA + std::min(ReadSize, (u32)0x100 + std::abs(DMAICounter / 4));
+	u32 buff1end = ActiveTSA + std::min(ReadSize, (u32)0x100 + std::abs(DMAICounter / 48));
 	u32 buff2end = 0;
 
 	if (buff1end > 0x100000)
@@ -426,9 +426,9 @@ void V_Core::FinishDMAread()
 
 	// DMA Reads are done AFTER the delay, so to get the timing right we need to scheule one last DMA to catch IRQ's
 	if (ReadSize)
-		DMAICounter = std::min(ReadSize, (u32)0x100) * 4;
+		DMAICounter = std::min(ReadSize, (u32)0x100) * 48;
 	else
-		DMAICounter = 4;
+		DMAICounter = 48;
 
 	CounterUpdate(DMAICounter);
 
@@ -446,7 +446,7 @@ void V_Core::DoDMAread(u16* pMem, u32 size)
 	ReadSize = size;
 	IsDMARead = true;
 	LastClock = psxRegs.cycle;
-	DMAICounter = std::min(ReadSize, (u32)0x100) * 4;
+	DMAICounter = (std::min(ReadSize, (u32)0x100) * 48);
 	Regs.STATX &= ~0x80;
 	Regs.STATX |= 0x400;
 	//Regs.ATTR |= 0x30;
@@ -470,7 +470,7 @@ void V_Core::DoDMAwrite(u16* pMem, u32 size)
 	{
 		Regs.STATX &= ~0x80;
 		//Regs.ATTR |= 0x30;
-		DMAICounter = 1 * 4;
+		DMAICounter = 1 * 48;
 		LastClock = psxRegs.cycle;
 		return;
 	}