Handle accumulator differently, handle segment registers, handle unary, add instructions

asm_files/list-0042.asm

@@ -0,0 +1,451 @@
+; ========================================================================
+;
+; (C) Copyright 2023 by Molly Rocket, Inc., All Rights Reserved.
+;
+; This software is provided 'as-is', without any express or implied
+; warranty. In no event will the authors be held liable for any damages
+; arising from the use of this software.
+;
+; Please see https://computerenhance.com for further information
+;
+; ========================================================================
+
+; ========================================================================
+; LISTING 42
+; ========================================================================
+
+;
+; NOTE(casey): This is not meant to be a real compliance test for 8086
+; disassemblers. It's just a reasonable selection of opcodes and patterns
+; to use as a first pass in making sure a disassembler handles a large
+; cross-section of the encoding. To be absolutely certain you haven't
+; missed something, you would need a more exhaustive listing!
+;
+
+bits 16
+
+mov si, bx
+mov dh, al
+mov cl, 12
+mov ch, -12
+mov cx, 12
+mov cx, -12
+mov dx, 3948
+mov dx, -3948
+mov al, [bx + si]
+mov bx, [bp + di]
+mov dx, [bp]
+mov ah, [bx + si + 4]
+mov al, [bx + si + 4999]
+mov [bx + di], cx
+mov [bp + si], cl
+mov [bp], ch
+mov ax, [bx + di - 37]
+mov [si - 300], cx
+mov dx, [bx - 32]
+mov [bp + di], byte 7
+mov [di + 901], word 347
+mov bp, [5]
+mov bx, [3458]
+mov ax, [2555]
+mov ax, [16]
+mov [2554], ax
+mov [15], ax
+
+push word [bp + si]
+push word [3000]
+push word [bx + di - 30]
+push cx
+push ax
+push dx
+push cs
+
+pop word [bp + si]
+pop word [3]
+pop word [bx + di - 3000]
+pop sp
+pop di
+pop si
+pop ds
+
+xchg ax, [bp - 1000]
+xchg [bx + 50], bp
+
+xchg ax, ax
+xchg ax, dx
+xchg ax, sp
+xchg ax, si
+xchg ax, di
+
+xchg cx, dx
+xchg si, cx
+xchg cl, ah
+
+in al, 200
+in al, dx
+in ax, dx
+
+out 44, ax
+out dx, al
+
+xlat
+lea ax, [bx + di + 1420]
+lea bx, [bp - 50]
+lea sp, [bp - 1003]
+lea di, [bx + si - 7]
+
+lds ax, [bx + di + 1420]
+lds bx, [bp - 50]
+lds sp, [bp - 1003]
+lds di, [bx + si - 7]
+
+les ax, [bx + di + 1420]
+les bx, [bp - 50]
+les sp, [bp - 1003]
+les di, [bx + si - 7]
+
+lahf
+sahf
+pushf
+popf
+
+add cx, [bp]
+add dx, [bx + si]
+add [bp + di + 5000], ah
+add [bx], al
+add sp, 392
+add si, 5
+add ax, 1000
+add ah, 30
+add al, 9
+add cx, bx
+add ch, al
+
+adc cx, [bp]
+adc dx, [bx + si]
+adc [bp + di + 5000], ah
+adc [bx], al
+adc sp, 392
+adc si, 5
+adc ax, 1000
+adc ah, 30
+adc al, 9
+adc cx, bx
+adc ch, al
+
+inc ax
+inc cx
+inc dh
+inc al
+inc ah
+inc sp
+inc di
+inc byte [bp + 1002]
+inc word [bx + 39]
+inc byte [bx + si + 5]
+inc word [bp + di - 10044]
+inc word [9349]
+inc byte [bp]
+
+aaa
+daa
+
+sub cx, [bp]
+sub dx, [bx + si]
+sub [bp + di + 5000], ah
+sub [bx], al
+sub sp, 392
+sub si, 5
+sub ax, 1000
+sub ah, 30
+sub al, 9
+sub cx, bx
+sub ch, al
+
+sbb cx, [bp]
+sbb dx, [bx + si]
+sbb [bp + di + 5000], ah
+sbb [bx], al
+sbb sp, 392
+sbb si, 5
+sbb ax, 1000
+sbb ah, 30
+sbb al, 9
+sbb cx, bx
+sbb ch, al
+
+dec ax
+dec cx
+dec dh
+dec al
+dec ah
+dec sp
+dec di
+dec byte [bp + 1002]
+dec word [bx + 39]
+dec byte [bx + si + 5]
+dec word [bp + di - 10044]
+dec word [9349]
+dec byte [bp]
+
+neg ax
+neg cx
+neg dh
+neg al
+neg ah
+neg sp
+neg di
+neg byte [bp + 1002]
+neg word [bx + 39]
+neg byte [bx + si + 5]
+neg word [bp + di - 10044]
+neg word [9349]
+neg byte [bp]
+
+cmp bx, cx
+cmp dh, [bp + 390]
+cmp [bp + 2], si
+cmp bl, 20
+cmp byte [bx], 34
+cmp ax, 23909
+
+aas
+das
+
+mul al
+mul cx
+mul word [bp]
+mul byte [bx + di + 500]
+
+imul ch
+imul dx
+imul byte [bx]
+imul word [9483]
+
+aam
+
+div bl
+div sp
+div byte [bx + si + 2990]
+div word [bp + di + 1000]
+
+idiv ax
+idiv si
+idiv byte [bp + si]
+idiv word [bx + 493]
+
+aad
+cbw
+cwd
+    
+not ah
+not bl
+not sp
+not si
+not word [bp]
+not byte [bp + 9905]
+
+shl ah, 1
+shr ax, 1
+sar bx, 1
+rol cx, 1
+ror dh, 1
+rcl sp, 1
+rcr bp, 1
+
+shl word [bp + 5], 1
+shr byte [bx + si - 199], 1
+sar byte [bx + di - 300], 1
+rol word [bp], 1
+ror word [4938], 1
+rcl byte [3], 1
+rcr word [bx], 1
+
+shl ah, cl
+shr ax, cl
+sar bx, cl
+rol cx, cl
+ror dh, cl
+rcl sp, cl
+rcr bp, cl
+
+shl word [bp + 5], cl
+shr word [bx + si - 199], cl
+sar byte [bx + di - 300], cl
+rol byte [bp], cl
+ror byte [4938], cl
+rcl byte [3], cl
+rcr word [bx], cl
+
+and al, ah
+and ch, cl
+and bp, si
+and di, sp
+and al, 93
+and ax, 20392
+and [bp + si + 10], ch
+and [bx + di + 1000], dx
+and bx, [bp]
+and cx, [4384]
+and byte [bp - 39], 239
+and word [bx + si - 4332], 10328
+
+test bx, cx
+test dh, [bp + 390]
+test [bp + 2], si
+test bl, 20
+test byte [bx], 34
+test ax, 23909
+
+or al, ah
+or ch, cl
+or bp, si
+or di, sp
+or al, 93
+or ax, 20392
+or [bp + si + 10], ch
+or [bx + di + 1000], dx
+or bx, [bp]
+or cx, [4384]
+or byte [bp - 39], 239
+or word [bx + si - 4332], 10328
+
+xor al, ah
+xor ch, cl
+xor bp, si
+xor di, sp
+xor al, 93
+xor ax, 20392
+xor [bp + si + 10], ch
+xor [bx + di + 1000], dx
+xor bx, [bp]
+xor cx, [4384]
+xor byte [bp - 39], 239
+xor word [bx + si - 4332], 10328
+
+rep movsb
+rep cmpsb
+rep scasb
+rep lodsb
+rep movsw
+rep cmpsw
+rep scasw
+rep lodsw
+
+; NOTE(casey): Special thanks (as always!) to Mārtiņš Možeiko for figuring out why NASM
+; wouldn't compile "rep stds" instructions. It was because it was a misprint in the 8086
+; manual! It was really just "rep stos", which of course is still in x64, and NASM
+; assembles it just fine.
+rep stosb
+rep stosw
+
+call [39201]
+call [bp - 100]
+call sp
+call ax
+
+jmp ax
+jmp di
+jmp [12]
+jmp [4395]
+    
+ret -7
+ret 500
+ret
+    
+label:
+je label
+jl label
+jle label
+jb label
+jbe label
+jp label
+jo label
+js label
+jne label
+jnl label
+jg label
+jnb label
+ja label
+jnp label
+jno label
+jns label
+loop label
+loopz label
+loopnz label
+jcxz label
+    
+int 13
+int3
+    
+into
+iret
+
+clc
+cmc
+stc
+cld
+std
+cli
+sti
+hlt
+wait
+
+lock not byte [bp + 9905]
+lock xchg [100], al
+
+mov al, cs:[bx + si]
+mov bx, ds:[bp + di]
+mov dx, es:[bp]
+mov ah, ss:[bx + si + 4]
+
+and ss:[bp + si + 10], ch
+or ds:[bx + di + 1000], dx
+xor bx, es:[bp]
+cmp cx, es:[4384]
+test byte cs:[bp - 39], 239
+sbb word cs:[bx + si - 4332], 10328
+
+lock not byte CS:[bp + 9905]
+
+;
+; NOTE(casey): These were not in the original homework set, but have been added since, as people
+; found instruction encodings that were not previously covered. Thank you to everyone who
+; submitted test cases!
+;
+
+call 123:456
+jmp 789:34
+
+mov [bx+si+59],es
+
+jmp 2620
+call 11804
+
+retf 17556
+ret 17560
+retf
+ret
+
+call [bp+si-0x3a]
+call far [bp+si-0x3a]
+jmp [di]
+jmp far [di]
+
+jmp 21862:30600
+
+;
+; TODO(casey): I would like to uncomment this, but as far as I can tell, NASM doesn't recognize the ESC instruction :(
+; so even if I just force the assembler to output the bits here, our disasm will fail to assemble because it will (correctly!)
+; print the esc instruction and NASM will error because it doesn't know what that is.
+;
+; esc 938,ax
+;
+
+;
+; TODO(casey): According to NASM, "rep movsb" is "not lockable". However the 8086 manual seems to think it is, and
+; even describes what happens when you you lock a rep: the lock is held for the duration of the rep operation. So...
+; yeah. Not sure why this doesn't work in NASM:
+;
+; lock rep movsb
+;

decoder8086.odin

@@ -20,6 +20,9 @@ Register :: struct {
 OpName :: enum {
     TBD,
     MOV,
+    PUSH,
+    POP,
+    XCHG,
     ADD,
     SUB,
     CMP,
@@ -69,6 +72,13 @@ registers := [8]Register {
     {fullname = "di", bytename = "bh", code = 0b111},
 }
 
+segment_registers := [4]Register {
+    {fullname = "es", code = 0b000},
+    {fullname = "cs", code = 0b001},
+    {fullname = "ss", code = 0b010},
+    {fullname = "ds", code = 0b011},
+}
+
 RegInfo :: struct {
     in_first_byte: bool,
     shift_offset: u8,
@@ -82,11 +92,61 @@ OpCodeId :: enum {
 
 LastBit :: struct{}
 FourthBit :: struct{}
+Force :: struct{}
 
 WordSize :: union {
     None,
     LastBit,
     FourthBit,
+    Force,
+}
+
+None :: struct {}
+
+Disp8 :: i8
+Disp16 :: i16
+Displacement :: union {
+    None,
+    Disp8,
+    Disp16
+}
+
+Value8 :: i8
+Value16 :: i16
+Data :: union {
+    None,
+    Value8,
+    Value16
+}
+
+ModMemory :: struct {}
+Mod8BitDisp :: i8
+Mod16BitDisp :: i16
+ModRegister :: struct {}
+
+ModMode :: union {
+    ModMemory,
+    Mod8BitDisp,
+    Mod16BitDisp,
+    ModRegister,
+}
+
+RegisterId :: distinct u8
+Immediate8 :: distinct i8
+Immediate16 :: distinct i16
+MemoryAddr :: struct {
+    addr_id: u8,
+    displacement: Displacement
+}
+Accumulator :: distinct i16
+SegmentRegister :: distinct i8
+OperandType :: union {
+    RegisterId,
+    Immediate8,
+    Immediate16,
+    MemoryAddr,
+    Accumulator,
+    SegmentRegister,
 }
 
 InstructionInfo :: struct {
@@ -99,17 +159,19 @@ InstructionInfo :: struct {
     reg_info: Maybe(RegInfo),
     has_data: bool,
     has_address: bool,
-    has_accumulator: bool,
+    uses_accumulator: bool,
     has_segreg: bool,
     has_flip: bool,
     has_sign_extension: bool,
     is_jump: bool,
+    is_unary: bool,
 }
 
 // TODO: Maybe we can get rid of it since I don't have to specify the shift_offset,
 //       not like it changes a lot
 reg_first_last := RegInfo{ in_first_byte = true, shift_offset = 0 }
 reg_second_middle := RegInfo{ in_first_byte = false, shift_offset = 3 }
+reg_first_middle := RegInfo{ in_first_byte = true, shift_offset = 3 }
 
 instructions := [?]InstructionInfo {
     { opname = .MOV, desc = "Register/memory to/from register", mask = 0b11111100, encoding = 0b10001000,
@@ -119,13 +181,29 @@ instructions := [?]InstructionInfo {
     { opname = .MOV, desc = "Immediate to register", mask = 0b11110000, encoding = 0b10110000,
       reg_info = reg_first_last, has_data = true, word_size = FourthBit{} },
     { opname = .MOV, desc = "Memory to accumulator", mask = 0b11111110, encoding = 0b10100000,
-      has_flip = true, word_size = LastBit{}, has_accumulator = true },
+      has_flip = true, word_size = LastBit{}, uses_accumulator = true },
     { opname = .MOV, desc = "Accumulator to memory", mask = 0b11111110, encoding = 0b10100010,
-      has_flip = true, word_size = LastBit{}, has_accumulator = true },
+      has_flip = true, word_size = LastBit{}, uses_accumulator = true },
     { opname = .MOV, desc = "Register/memory to segment register", mask = 0b11111111, encoding = 0b10001110,
       has_segreg = true, has_address = true, word_size = None{} },
     { opname = .MOV, desc = "Segment register to register/memory", mask = 0b11111111, encoding = 0b10001100,
       has_segreg = true, has_address = true, word_size = None{} },
+    { opname = .PUSH, desc = "", mask = 0b11111111, encoding = 0b11111111,
+      has_address = true, word_size = None{}, is_unary = true },
+    { opname = .PUSH, desc = "", mask = 0b11111000, encoding = 0b01010000,
+      reg_info = reg_first_last, word_size = Force{}, is_unary = true },
+    { opname = .PUSH, desc = "", mask = 0b11100111, encoding = 0b00000110,
+      has_segreg = true, reg_info = reg_first_middle, word_size = Force{}, is_unary = true },
+    { opname = .POP, desc = "", mask = 0b11111111, encoding = 0b10001111,
+      has_address = true, word_size = None{}, is_unary = true },
+    { opname = .POP, desc = "", mask = 0b11111000, encoding = 0b01011000,
+      reg_info = reg_first_last, word_size = Force{}, is_unary = true },
+    { opname = .POP, desc = "", mask = 0b11100111, encoding = 0b00000111,
+      has_segreg = true, reg_info = reg_first_middle, word_size = None{}, is_unary = true },
+    { opname = .XCHG, desc = "", mask = 0b11111110, encoding = 0b10000110,
+      reg_info = reg_second_middle, has_address = true, word_size = LastBit{}, has_flip = true},
+    { opname = .XCHG, desc = "", mask = 0b11111000, encoding = 0b10010000,
+      reg_info = reg_first_last, uses_accumulator = true, word_size = Force{}, },
     { opname = .TBD, desc = "Reg/memory with register to either", mask = 0b11000100, encoding = 0b00000000,
       opcode_id = .First, reg_info = reg_second_middle, has_address = true, word_size = LastBit{}, has_flip = true },
     { opname = .TBD, desc = "Immediate to register/memory", mask = 0b11111100, encoding = 0b10000000,
@@ -182,52 +260,6 @@ instructions := [?]InstructionInfo {
     { opname = .JCXZ,   desc = "Jump on not zero", mask = 0b11111111, encoding = 0b11100011, is_jump = true},
 }
 
-None :: struct {}
-
-Disp8 :: i8
-Disp16 :: i16
-Displacement :: union {
-    None,
-    Disp8,
-    Disp16
-}
-
-Value8 :: i8
-Value16 :: i16
-Data :: union {
-    None,
-    Value8,
-    Value16
-}
-
-ModMemory :: struct {}
-Mod8BitDisp :: i8
-Mod16BitDisp :: i16
-ModRegister :: struct {}
-
-ModMode :: union {
-    ModMemory,
-    Mod8BitDisp,
-    Mod16BitDisp,
-    ModRegister,
-}
-
-RegisterId :: distinct u8
-Immediate8 :: distinct i8
-Immediate16 :: distinct i16
-MemoryAddr :: struct {
-    addr_id: u8,
-    displacement: Displacement
-}
-Accumulator :: distinct i16
-OperandType :: union {
-    RegisterId,
-    Immediate8,
-    Immediate16,
-    MemoryAddr,
-    Accumulator,
-}
-
 inst_map := make(map[u8]InstructionInfo)
 RIGHT_ALIGN_AMOUNT := 35
 
@@ -285,6 +317,8 @@ get_memory_type_string :: proc(mem_type: OperandType, is_word: bool) -> string {
         string_val = get_memory_string(val)
     case Accumulator:
         string_val = fmt.aprintf("[%d]", val)
+    case SegmentRegister:
+        string_val = segment_registers[val].fullname
     }
     return string_val
 }
@@ -352,7 +386,7 @@ main :: proc() {
     }
     defer os.close(f)
 
-    data := make([]u8, 512)
+    data := make([]u8, 1024)
     bytes_read, err2 := os.read(f, data)
     if err2 != nil {
         // ...
@@ -369,13 +403,17 @@ main :: proc() {
     // asdf :u16 = 0b00000011_11101000
     // asdf2 :i16 = (i16)(asdf)
     // fmt.printfln("%d", asdf2)
+    print_at_end := false
     read_next := false
     src_dst := true
     idx := 0
     added_label := false
     line_count := 0
+    // last_opname: string
+    last_opname: [3]byte
     instruction_builder := strings.builder_make()
-    instruction_list := make([dynamic]string, 128)
+    instruction_list := make([dynamic]string, 512)
+    fmt.println("bits 16")
     for idx < bytes_read {
         processed := 1
         curr_byte := data[idx]
@@ -408,6 +446,7 @@ main :: proc() {
         switch val in instruction.word_size {
         case LastBit:   is_word = curr_byte & 1 == 1
         case FourthBit: is_word = curr_byte & 0b0000_1000 != 0
+        case Force:     is_word = true
         case None:
         }
 
@@ -445,6 +484,10 @@ main :: proc() {
             } else if mod == 3 {
                 lhs2 = (RegisterId)(registers[rm].code)
             }
+        } else if instruction.has_segreg {
+            lhs2 = (SegmentRegister)(segment_registers[reg].code)
+        } else if instruction.uses_accumulator {
+            lhs2 = (RegisterId)(registers[0].code)
         } else {
             lhs2 = (RegisterId)(registers[reg].code)
         }
@@ -456,12 +499,16 @@ main :: proc() {
             processed += word_signed ? 2 : 1
             rhs2 = (OperandType)(word_signed ? (Immediate16)(get_i16(data[data_idx:])) : (Immediate8)(data[data_idx]))
             has_immediate = true
-        } else if instruction.has_accumulator {
+        } else if instruction.uses_accumulator {
+            if _, ok := instruction.word_size.(LastBit); ok {
                 processed += is_word ? 2 : 1
                 rhs2 = (OperandType)(is_word ? (Accumulator)(get_i16(data[data_idx:])) : (Accumulator)(data[data_idx]))
             } else {
                 rhs2 = (RegisterId)(reg)
             }
+        } else {
+            rhs2 = (RegisterId)(reg)
+        }
 
         if flip_dst {
             lhs2, rhs2 = rhs2, lhs2
@@ -486,6 +533,11 @@ main :: proc() {
             value := (i8)(data[idx+1]) + 2
             full_inst = fmt.aprintf("%s $%s%d   ; %d", strings.to_lower(opname), value >= 0 ? "+" : "", value, value - 2)
             processed += 1
+        } else if instruction.is_unary {
+            if instruction.has_address {
+                size_string = "word "
+            }
+            full_inst = fmt.aprintf("%s %s%s", opname, size_string, lhs)
         } else {
             opname = strings.to_lower(opname)
             if opname == "mov" {
@@ -498,13 +550,28 @@ main :: proc() {
         for i in 0..<processed {
             fmt.sbprintf(&instruction_builder, " %08b", data[idx + i])
         }
+        if print_at_end {
             instruction_list[line_count] = strings.clone(strings.to_string(instruction_builder))
+        } else {
+            op := strings.to_string(instruction_builder)
+            if op[0:3] != string(last_opname[:]) {
+                fmt.println()
+            }
+            copy(last_opname[:], op[0:3])
+            fmt.println(op)
+        }
         idx += processed
         line_count += 1
         strings.builder_reset(&instruction_builder)
     }
-    fmt.println("bits 16\n")
+    if print_at_end {
         for i in 0..<line_count {
+            opname := instruction_list[i]
+            if !strings.has_prefix(opname, string(last_opname[:])) {
+                fmt.println()
+            }
+            copy(last_opname[:], opname[0:3])
             fmt.println(instruction_list[i])
         }
+    }
 }