Wednesday, April 18, 2018

Loading stuff into registers


The bulk of the work that a processor does is done through registers. The typical pattern is that you load a value into a register, manipulate it in some way, then store the data back to memory, possibly in a different location. As was discussed in a much earlier article, the LDA, LDX, and LDY instructions for loading memory do so with a variety of address modes. These address modes determine which address is to be accessed. Zero page is easy to deal with as the next byte is the address that is being addressed. The zero page index modes likewise use the next byte as the address but then add the values of the x or y register to this value to come up with the final address.

Absolute addressing does need to manipulate two bytes to determine the address and is very commonly used so there is a utility function that uses the address of the instruction and takes the two bytes after the instruction to come up with the address. Absolute index modes also use this function but then add the values of the x or y register. This function can be used for non-instructions by having the address be one less than the desired first byte of the address, which is a trick that I will be using.

    private fun findAbsoluteAddress(address:Int):Int {
        return (mem.read(address+2) * 256 + mem.read(address+1))
    }

As the load instructions and many other instructions we will be implementing in the future need to set flags, I wrote a quick utility function for handling the setting of a flag bit to a Boolean value. Based on whether the bit is being set or cleared the method will use the appropriate Boolean operation to perform the bit setting or clearing. Setting is a simple or function while clearing requires creation of an inverse mask which then gets ANDed with the flag byte.

    private fun adjustFlag(flag:Int, isSet:Boolean) {
        if (isSet)
            state.flags = state.flags or flag
        else
            state.flags = state.flags and (255 xor flag)
    }

There is also some timing issues that we need to deal with when a page boundary is crossed with the indirect addressing functions. Timing is very important to an emulator, especially one for the 2600 as each scan line has 76 cycles for manipulating the TIA chip so precice processor timing is a necessity to properly emulate the machine.

    private fun pageBoundsCheck(baseAddress:Int, targetAddress:Int) {
        val basePage =  baseAddress / 256
        val actualPage = targetAddress / 256
        if (basePage != actualPage)
            ++state.tick
    }

Which leaves us with a general function for loading a byte from the specified address with the indicated offset. This function is called by the LDA, LDX, and LDY methods setting the parameters to the appropriate base address which was determined by the address mode. If the address mode is one that has an offset value then that offset is included. Finally, for address modes that incur extra cycles on page bound violations, we include a flag to indicate that this needs to be taken into account.

    fun loadByteFromAddress(address:Int, offset:Int = 0, checkBounds:Boolean = false):Int {
        val result:Int = mem.read(address + offset)
        adjustFlag (ZERO_FLAG, result == 0)
        adjustFlag (NEGATIVE_FLAG, (result and 128) > 0)
        if (checkBounds)
            pageBoundsCheck(address, address+offset)

        return result
    }

Here are the LDA, LDX, and LDY instructions that were implemented.

LDALoaD Accumulator
Address Mode
Decimal OPCode
Hexadecimal OpCode
Size
Cycles
#Immediate
169
$A9
2
2
Zero Page
165
$A5
2
3
Zero Page,X
181
$B5
2
4
Absolute
173
$AD
3
4
Absolute,X
189
$BD
3
4-5
Absolute,Y
185
$B9
3
4-5
(Indirect, X)
161
$A1
2
6
(Indirect),Y
177
$B1
2
5-6
Flags affected: NZ
Cycle Notes: Indirect modes take extra cycle if page boundaries crossed
Usage: Loads the accumulator with a value stored in memory. Used for manipulating memory, especially if math or Boolean operations need to be performed on that memory.
Test Code:
; LDA Immediate - zero test
     LDA #0
     BRK
     ; expect a=0; z=1; n=0
; LDA (Indirect,X)
     LDX #200
     LDA (54,X)
     BRK
.ORG 254
.WORD 1024
.ORG 1024
.BYTE 88
     ; expect A=88
    
    
; LDA (Indirect),Y
     LDY #10
     LDA (254),Y
     BRK
.ORG 254
.WORD 1024
.ORG 1034
.BYTE 88
     ; expect A=88

Implementation:
// immediate mode
m.state.acc = m.loadByteFromAddress(m.state.ip, 1)
// (indirect,x)
m.state.acc = m.loadByteFromAddress(m.findAbsoluteAddress(
           // need to deduct 1 from loadByteFromAddress method as it is designed to skip opcode
           ((m.loadByteFromAddress(m.state.ip, 1)+m.state.x) and 255)-1))
// (indirect),y
m.state.acc = m.loadByteFromAddress(m.findAbsoluteAddress(
     // need to deduct 1 from loadByteFromAddress method as it is designed to skip opcode
     (m.loadByteFromAddress(m.state.ip, 1)-1)), m.state.y, true)

LDXLoaD index X
Address Mode
Decimal OPCode
Hexadecimal OpCode
Size
Cycles
#Immediate
162
$A2
2
2
Zero Page
166
$A6
2
3
Zero Page,Y
182
$B6
2
4
Absolute
174
$AE
3
4
Aboslute,Y
190
$BE
3
4-5
Flags affected: NZ
Cycle Notes: Indexed modes take extra cycle if page boundaries crossed
Usage: Loads the X register with a value stored in memory. This is handy for setting offsets and counting values which are what the X register generally is used for.
Test Code:
; LDX Zero Page 0
     LDX 200
     BRK
.ORG 200
.BYTE 0
     ; expect x=0, z=1, n=0

; LDX Zero Page, y
     LDY #15
     LDX 10,X
     BRK
.ORG 25
.BYTE 52
     ; expect x=52

; LDA and LDX Absolute,Y
     LDY #10
     LDA 512,Y
     LDX 513,Y
     BRK
.ORG 522
.BYTE 1 2
     ; expect a = 1, x=2
Implementation:
// ldx – zero page
m.state.x = m.loadByteFromAddress(m.state.ip, 1)
// ldx zero page,y
m.state.x = m.loadByteFromAddress(m.loadByteFromAddress(m.state.ip, 1), m.state.y)
// LDX absolute,Y
m.state.x = m.loadByteFromAddress(m.findAbsoluteAddress(m.state.ip), m.state.y, true)


LDYLoaD index Y
Address Mode
Decimal OPCode
Hexadecimal OpCode
Size
Cycles
#immediate
160
$A0
2
2
Zero Page
164
$A4
2
3
Zero Page,X
180
$B4
2
4
Absolute
172
$AC
3
4
Absolute,X
188
$BC
3
4-5
Flags affected: NZ
Cycle Notes: Indexed modes take extra cycle if page boundaries crossed
Usage: Loads the Y register with a value stored in memory. This is handy for setting offsets and counting values which are what the Y register generally is used for.
Test Code:
; LDY Zero Page, x
     LDX #5
     LDY 5,X
     BRK
.ORG 10
.BYTE 12
     ; expect y=12

; LDA, LDX, LDY Absolute
     LDA 1024
     LDX 1025
     LDY 1026
     BRK
.ORG 1024
.BYTE 1 2 3
     ; expect a = 1, x=2, y=3

; LDA and LDY Absolute,X
     LDX #10
     LDA 512,X
     LDY 513,X
     BRK
.ORG 522
.BYTE 1 2
     ; expect a = 1, y=2  
Implementation:
// LDY zero page, x
m.state.y = m.loadByteFromAddress(m.loadByteFromAddress(m.state.ip, 1), m.state.x)
// LDY absolute
m.state.y = m.loadByteFromAddress(m.findAbsoluteAddress(m.state.ip))
// LDY absolute,Y
m.state.y = m.loadByteFromAddress(m.findAbsoluteAddress(m.state.ip), m.state.x, true)

No comments:

Post a Comment