nulldc-360/nullDC/dc/pvr/pvrLock.cpp

/*
	Locking code for pvr plugins
	Was never finished
	Works for 64bit area locks
*/
#include "plugins/plugin_manager.h"
#include "pvrLock.h"
 
using namespace std;

vector<vram_block*> VramLocks[VRAM_SIZE/PAGE_SIZE];
//vram 32-64b
VArray2 vram;
 

//Address space convertion functions
void vramlock_Read32b(u32* pdst,u32 offset,u32 len)
{
	//TODO:implement .. later
}
 
//mmx version , 2x64bit at a time
//pdst needs to be 16 byte aligned
void vramlock_Read32b_mmx(u32* pdst,u32 offset,u32 len)
{
		//TODO:implement .. later
}
 
//sse version , 2x128bit at a time
//pdst needs to be 16 byte aligned
void vramlock_Read32b_sse(u32* pdst,u32 offset,u32 len)
{
		//TODO:implement .. later
}
 
 
 
//So, the bytes 0xA4000000-0xA4000003 correspond to 0xA5000000-0xA5000003, 0xA4000004-0xA4000007 to 0xA5400000-0xA5400003,
//0xA4000008-0xA400000B to 0xA5000004-0xA5000007 and so on. 
//
//
//Convert Sh4 address to vram_32 offset
 
//Convert offset32 to offset64
u32 vramlock_ConvOffset32toOffset64(u32 offset32)
{
		//64b wide bus is archevied by interleaving the banks every 32 bits
		//so bank is Address<<3
		//bits <4 are <<1 to create space for bank num
		//bank 0 is mapped at 400000 (32b offset) and after
		const u32 bank_bit=VRAM_MASK-(VRAM_MASK/2);
		const u32 static_bits=(VRAM_MASK-(bank_bit*2)+1)|3;
		const u32 moved_bits=VRAM_MASK-static_bits-bank_bit;

		u32 bank=(offset32&bank_bit)/bank_bit*4;//bank will be used as uper offset too
		u32 lv=offset32&static_bits; //these will survive
		offset32&=moved_bits;
		offset32<<=1;
		//       |inbank offset    |       bank id        | lower 2 bits (not changed)
		u32 rv=  offset32 + bank                  + lv;
 
		return rv;
}
//Convert Sh4 address to vram_64 offset
u32 vramlock_ConvAddrtoOffset64(u32 Address)
{
	if (Is_64_Bit(Address))
	{
		return  Address & VRAM_MASK;//64 bit offset
	}
	else
	{
		return vramlock_ConvOffset32toOffset64(Address);
	}
}
//List functions
//
void vramlock_list_remove(vram_block* block)
{
	u32 base = block->start >> PAGE_SHIFT;
	u32 end = block->end >> PAGE_SHIFT;

	for (u32 i=base;i<=end;i++)
	{
		vector<vram_block*>* list=&VramLocks[i];
		for (size_t j=0;j<list->size();j++)
		{
			if ((*list)[j]==block)
			{
				(*list)[j]=0;
			}
		}
	}
}
 
void vramlock_list_add(vram_block* block)
{
	u32 base = block->start >> PAGE_SHIFT;
	u32 end = block->end >> PAGE_SHIFT;


	for (u32 i=base;i<=end;i++)
	{
		vector<vram_block*>* list=&VramLocks[i];
		for (u32 j=0;j<list->size();j++)
		{
			if ((*list)[j]==0)
			{
				(*list)[j]=block;
				goto added_it;
			}
		}

		list->push_back(block);
added_it:
		__noop;
	}
}
 
//simple IsInRange test
inline bool IsInRange(vram_block* block,u32 offset)
{
	return (block->start<=offset) && (block->end>=offset);
}

//returns block handle
vram_block* FASTCALL vramlock_Lock_32(u32 start_offset32,u32 end_offset32,void* userdata)
{
	vram_block* block=(vram_block*)malloc(sizeof(vram_block));
 
	block->end=end_offset32;
	block->start=start_offset32;
	block->len=end_offset32-start_offset32+1;
	block->userdata=userdata;
	block->type=32;
 
	//okz , new idea
	//asuming locks are 4 byte alligned (textures have to be 4 byte aligned too , dont they ? 
	//									 anyhow , fb is 32b alligned and locks are 4kb aligned 
	//									 so it doesnt matter)
	//32b to 64b maping goes 2:1
	//so we can do
	//lock(bank1,size>>1);lock(bank2,size>>1);

	/*for (u32 base=start_offset32;base<end_offset32;base+=4)
	{
		SetVramTest_32(vramlock_ConvOffset32toOffset64(base));
	}
 
	for (u32 base=block->start;base<block->end;base+=4)
	{
		vramlock_list_add(block,&vram_locks[GetHash(vramlock_ConvOffset32toOffset64(base))]);
	}*/
 
	return block;
}

vram_block* FASTCALL vramlock_Lock_64(u32 start_offset64,u32 end_offset64,void* userdata)
{
	vram_block* block=(vram_block* )malloc(sizeof(vram_block));
 

	if (end_offset64>(VRAM_SIZE-1))
	{
		msgboxf("vramlock_Lock_64: end_offset64>(VRAM_SIZE-1) \n Tried to lock area out of vram , possibly bug on the pvr plugin",MBX_OK);
		end_offset64=(VRAM_SIZE-1);
	}

	if (start_offset64>end_offset64)
	{
		msgboxf("vramlock_Lock_64: start_offset64>end_offset64 \n Tried to lock negative block , possibly bug on the pvr plugin",MBX_OK);
		start_offset64=0;
	}

	

	block->end=end_offset64;
	block->start=start_offset64;
	block->len=end_offset64-start_offset64+1;
	block->userdata=userdata;
	block->type=64;

	vram.LockRegion(block->start,block->len);
	vramlock_list_add(block);

	return block;
}
 
bool VramLockedWrite(u8* address)
{
	size_t offset=address-vram.data;

	if (offset<VRAM_SIZE)
	{
		size_t addr_hash = offset >> PAGE_SHIFT;
		vector<vram_block*>* list=&VramLocks[addr_hash];
			
		for (size_t i=0;i<list->size();i++)
		{
			if ((*list)[i])
			{
				libPvr.LockedBlockWrite((*list)[i],(u32)offset);
				
				if ((*list)[i])
				{
					msgboxf("Error , pvr is suposed to remove lock",MBX_OK);
					__debugbreak(); 
				}

			}
		}
		list->clear();
		vram.UnLockRegion((u32)offset&(~(PAGE_SIZE-1)),PAGE_SIZE);

		return true;
	}
	else
		return false;
}

//unlocks mem
//also frees the handle
void FASTCALL vramlock_Unlock_block(vram_block* block)
{
	//VRAM_SIZE/PAGE_SIZE;
	if (block->end>VRAM_SIZE)
		msgboxf("Error : block end is after vram , skiping unlock",MBX_OK);
	else
	{
		vramlock_list_remove(block);
		//more work needed
		free(block);
	}
}