※ 引述《unixxxx (皓皓)》之銘言：
: 標題: Re: [討論] 寫三元判斷式code review被打槍
: 時間: Sat Dec 17 03:51:38 2022
: 很多Javascript 高手都是用 switch 取代
"特定"情況下的確是好方式
舉個例子 以前我在調校能時候有用過這種方式 這是c#的code部分節錄
void Mem_w(ushort address, byte value)
{
if (address < 0x2000) NES_MEM[address & 0x7ff] = value;
else if (address < 0x4020) IO_write(address, value);
else if (address < 0x6000) MapperRouterW_ExpansionROM(address,value);
else if (address < 0x8000) MapperRouterW_RAM(address, value);
else MapperRouterW_PRG(address, value);
}
這個是一個非常頻繁被呼叫的method 如果address 大於等於0x8000
其實前面就浪費好幾次計算在判斷式上
因此改成 功能上是相等的
static void Mem_w(ushort address, byte value)
{
switch (address & 0xf000)
{
case 0:
case 0x1000:
NES_MEM[address & 0x7ff] = value;
break;
case 0x2000:
case 0x3000:
case 0x4000:
IO_write(address, value);
break;
case 0x5000:
MapperObj.MapperW_ExpansionROM(address, value);
break;
case 0x6000:
case 0x7000:
MapperObj.MapperW_RAM(address, value);
break;
case 0x8000:
case 0x9000:
case 0xa000:
case 0xb000:
case 0xc000:
case 0xd000:
case 0xe000:
case 0xf000:
MapperObj.MapperW_PRG(address, value);
break;
}
}
但再簡化快速下去還有大招
初始化執行一次就好
static void init_function()
{
mem_write_fun = new Action<ushort, byte>[0x10000];
for (int address = 0; address < 0x10000; address++)
{
if (address < 0x2000) mem_write_fun[address] =
new Action<ushort, byte>((addr, val) => { NES_MEM[addr & 0x7ff] = val; });
else if (address < 0x4020) mem_write_fun[address] =
new Action<ushort,byte>(IO_write);
else if (address < 0x6000) mem_write_fun[address] =
new Action<ushort, byte>(MapperObj.MapperW_ExpansionROM);
else if (address < 0x8000) mem_write_fun[address] =
new Action<ushort, byte>(MapperObj.MapperW_RAM);
else mem_write_fun[address] =
new Action<ushort, byte>(MapperObj.MapperW_PRG);
}
}
之後讀取
[MethodImpl(MethodImplOptions.AggressiveInlining)]
static void Mem_w(ushort address, byte value)
{
mem_write_fun[address](address, value);
}
實際上跟switch原理很相似 但直接把mapper的實作 哪個address對應啥動作
直接寫到記憶址array去了
我是不知道你所謂的javascript高手用switch替代判斷式是碰到啥議題
但特定議題用改用switch 甚至自己刻死路由效能是會快非常多
以前有筆記在這邊...
https://dotblogs.com.tw/enet/2017/01/21/emu_optimization_1