Improved PPU Interpreter

- Fixed MULHW & MULHWU.
- Improved MULHD & MULHDU.

rpcs3/Emu/Cell/PPUInterpreter.h

@@ -6,6 +6,7 @@
 #include "Emu/SysCalls/SysCalls.h"
 #include "rpcs3.h"
 #include <stdint.h>
+#include <intrin.h>
 
 #define UNIMPLEMENTED() UNK(__FUNCTION__)
 
@@ -2272,6 +2273,11 @@ private:
 	}
 	void MULHDU(u32 rd, u32 ra, u32 rb, bool rc)
 	{
+#ifdef _M_X64
+		CPU.GPR[rd] = __umulh(CPU.GPR[ra], CPU.GPR[rb]);
+#else
+		ConLog.Warning("MULHDU");
+
 		const u64 RA = CPU.GPR[ra];
 		const u64 RB = CPU.GPR[rb];
 
@@ -2292,12 +2298,15 @@ private:
 		mid += (a0 + a1) * (b0 + b1) - (lo + hi);
 
 		CPU.GPR[rd] = RD._u64[1];
+#endif
 
 		if(rc) CPU.UpdateCR0<s64>(CPU.GPR[rd]);
 	}
 	void MULHWU(u32 rd, u32 ra, u32 rb, bool rc)
 	{
-		CPU.GPR[rd] = (CPU.GPR[ra] * CPU.GPR[rb]) >> 32;
+		u32 a = CPU.GPR[ra];
+		u32 b = CPU.GPR[rb];
+		CPU.GPR[rd] = ((u64)a * (u64)b) >> 32;
 		if(rc) CPU.UpdateCR0<s32>(CPU.GPR[rd]);
 	}
 	void MFOCRF(u32 a, u32 rd, u32 crm)
@@ -2344,9 +2353,9 @@ private:
 	}
 	void SLW(u32 ra, u32 rs, u32 rb, bool rc)
 	{
-		const u32 n = CPU.GPR[rb] & 0x1f;
+		u32 n = CPU.GPR[rb] & 0x1f;
-		const u32 r = rotl32((u32)CPU.GPR[rs], n);
+		u32 r = rotl32((u32)CPU.GPR[rs], n);
-		const u32 m = (CPU.GPR[rb] & 0x20) ? 0 : rotate_mask[32][63 - n];
+		u32 m = (CPU.GPR[rb] & 0x20) ? 0 : rotate_mask[32][63 - n];
 
 		CPU.GPR[ra] = r & m;
 
@@ -2357,18 +2366,18 @@ private:
 		u32 i;
 		for(i=0; i < 32; i++)
 		{
-			if(CPU.GPR[rs] & (0x80000000 >> i)) break;
+			if(CPU.GPR[rs] & (1ULL << (31 - i))) break;
 		}
 
 		CPU.GPR[ra] = i;
 
-		if(rc) CPU.UpdateCR0<s32>(CPU.GPR[ra]);
+		if(rc) CPU.SetCRBit(CR_LT, false);
 	}
 	void SLD(u32 ra, u32 rs, u32 rb, bool rc)
 	{
-		const u32 n = CPU.GPR[rb] & 0x3f;
+		u32 n = CPU.GPR[rb] & 0x3f;
-		const u64 r = rotl64(CPU.GPR[rs], n);
+		u64 r = rotl64(CPU.GPR[rs], n);
-		const u64 m = (CPU.GPR[rb] & 0x40) ? 0 : rotate_mask[0][63 - n];
+		u64 m = (CPU.GPR[rb] & 0x40) ? 0 : rotate_mask[0][63 - n];
 
 		CPU.GPR[ra] = r & m;
 
@@ -2435,15 +2444,14 @@ private:
 	}
 	void CNTLZD(u32 ra, u32 rs, bool rc)
 	{
-		u32 i = 0;
+		u32 i;
-			
+		for(i=0; i < 64; i++)
-		for(u64 mask = 1ULL << 63; i < 64; i++, mask >>= 1)
 		{
-			if(CPU.GPR[rs] & mask) break;
+			if(CPU.GPR[rs] & (1ULL << (63 - i))) break;
 		}
 
 		CPU.GPR[ra] = i;
-		if(rc) CPU.UpdateCR0<u64>(CPU.GPR[ra]);
+		if(rc) CPU.SetCRBit(CR_LT, false);
 	}
 	void ANDC(u32 ra, u32 rs, u32 rb, bool rc)
 	{
@@ -2459,6 +2467,10 @@ private:
 	}
 	void MULHD(u32 rd, u32 ra, u32 rb, bool rc)
 	{
+#ifdef _M_X64
+		CPU.GPR[rd] = __mulh(CPU.GPR[ra], CPU.GPR[rb]);
+#else
+		ConLog.Warning("MULHD");
 		const s64 RA = CPU.GPR[ra];
 		const s64 RB = CPU.GPR[rb];
 
@@ -2479,13 +2491,16 @@ private:
 		mid += (a0 + a1) * (b0 + b1) - (lo + hi);
 
 		CPU.GPR[rd] = RT._u64[1];
+#endif
 
 		if(rc) CPU.UpdateCR0<s64>(CPU.GPR[rd]);
 	}
 	void MULHW(u32 rd, u32 ra, u32 rb, bool rc)
 	{
-		CPU.GPR[rd] = (s64)(s32)((CPU.GPR[ra] * CPU.GPR[rb]) >> 32);
+		s32 a = CPU.GPR[ra];
-		if(rc) CPU.UpdateCR0<s64>(CPU.GPR[rd]);
+		s32 b = CPU.GPR[rb];
+		CPU.GPR[rd] = ((s64)a * (s64)b) >> 32;
+		if(rc) CPU.UpdateCR0<s32>(CPU.GPR[rd]);
 	}
 	void LDARX(u32 rd, u32 ra, u32 rb)
 	{
@@ -2889,8 +2904,8 @@ private:
 	void SRW(u32 ra, u32 rs, u32 rb, bool rc)
 	{
 		u32 n = CPU.GPR[rb] & 0x1f;
-		u64 r = rotl32((u32)CPU.GPR[rs], 64 - n);
+		u32 r = rotl32((u32)CPU.GPR[rs], 64 - n);
-		u64 m = (CPU.GPR[rb] & 0x20) ? 0 : rotate_mask[32 + n][63];
+		u32 m = (CPU.GPR[rb] & 0x20) ? 0 : rotate_mask[32 + n][63];
 		CPU.GPR[ra] = r & m;
 		
 		if(rc) CPU.UpdateCR0<s32>(CPU.GPR[ra]);

rpcs3/Emu/GS/GL/GLGSRender.cpp

@@ -124,16 +124,21 @@ void GLGSRender::EnableVertexData(bool indexed_draw)
 	static u32 offset_list[m_vertex_count];
 	u32 cur_offset = 0;
 
+	const u32 data_offset = indexed_draw ? 0 : m_draw_array_first;
+
 	for(u32 i=0; i<m_vertex_count; ++i)
 	{
 		offset_list[i] = cur_offset;
 
 		if(!m_vertex_data[i].IsEnabled() || !m_vertex_data[i].addr) continue;
 
-		cur_offset += m_vertex_data[i].data.GetCount();
+		const size_t item_size = m_vertex_data[i].GetTypeSize() * m_vertex_data[i].size;
+		const size_t data_size = m_vertex_data[i].data.GetCount() - data_offset * item_size;
 		const u32 pos = m_vdata.GetCount();
-		m_vdata.InsertRoomEnd(m_vertex_data[i].data.GetCount());
+
-		memcpy(&m_vdata[pos], &m_vertex_data[i].data[0], m_vertex_data[i].data.GetCount());
+		cur_offset += data_size;
+		m_vdata.InsertRoomEnd(data_size);
+		memcpy(&m_vdata[pos], &m_vertex_data[i].data[data_offset * item_size], data_size);
 	}
 
 	m_vao.Create();
@@ -1010,7 +1015,7 @@ void GLGSRender::ExecCMD()
 	if(m_draw_array_count)
 	{
 		//ConLog.Warning("glDrawArrays(%d,%d,%d)", m_draw_mode - 1, m_draw_array_first, m_draw_array_count);
-		glDrawArrays(m_draw_mode - 1, m_draw_array_first, m_draw_array_count);
+		glDrawArrays(m_draw_mode - 1, 0, m_draw_array_count);
 		checkForGlError("glDrawArrays");
 		DisableVertexData();
 	}

rpcs3/Emu/GS/RSXThread.cpp

@@ -499,6 +499,7 @@ void RSXThread::DoCmd(const u32 fcmd, const u32 cmd, mem32_ptr_t& args, const u3
 		const u32 addr = GetAddress(args[0] & 0x7fffffff, args[0] >> 31);
 		CMD_LOG("num=%d, addr=0x%x", index, addr);
 		m_vertex_data[index].addr = addr;
+		m_vertex_data[index].data.ClearF();
 	}
 	break;