JosephFerano/performance-aware
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

WIP Refactor: Instruction parser now works and basic decoding in place

Browse Source
This commit is contained in:
Joseph Ferano 2024-01-16 10:09:50 +07:00
parent 4754a8cd4c
commit 8497316768
2 changed files with 107 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

131
decode.c
View File

@ -3,8 +3,6 @@
#include "lib.h" #include "lib.h"
#include <sys/stat.h> #include <sys/stat.h>
#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]))
enum InstructionType enum InstructionType
{ {
INST_MOV_REG_REG = 0b10001000, INST_MOV_REG_REG = 0b10001000,
@ -92,9 +90,19 @@ static i16 get_data(unsigned char* buf, char wide)
// Cast buf[0] to sbyte if not the conversion to i16 won't detect signedness // Cast buf[0] to sbyte if not the conversion to i16 won't detect signedness
return wide == 1 ? (i16)buf[1] << 8 | buf[0] : (sbyte)buf[0]; return wide == 1 ? (i16)buf[1] << 8 | buf[0] : (sbyte)buf[0];
} }
#define IS_INST(value, m, inst) ((value & ~m) == (char)inst)
typedef struct InstMask {char *name; u8 mask; u8 code;} InstMask; static u8 mask_and_shift(u8 value, u8 mask)
{
value &= mask;
int count = 0;
while ((mask & 0x1) == 0 && count < 8)
{
value >>= 1;
mask >>= 1;
count++;
}
return value;
}
typedef struct InstFormat typedef struct InstFormat
{ {
@ -103,7 +111,7 @@ typedef struct InstFormat
u8 inst_enc; u8 inst_enc;
u8 mask_inst; u8 mask_inst;
u8 mask_w; u8 mask_w;
u64 mask_reg; u8 mask_reg;
bool has_operands; bool has_operands;
bool has_displacement; bool has_displacement;
bool has_data; bool has_data;
@ -122,14 +130,15 @@ InstFormat inst_formats[] =
// MOV // MOV
//////// ////////
// Register/memory to/from register // Register/memory to/from register
{.id=1, .name="mov", .inst_enc=0b10001000, .mask_inst=0x3, .has_operands=true, .has_displacement=true, {.id=1, .name="mov", .inst_enc=0b10001000, .mask_inst=0x3, .mask_reg=0b00111000,
.has_d=true, .has_w=true, .has_reg=true, .has_mod=true, .has_rm=true}, .mask_w=0x1, .has_operands=true, .has_displacement=true, .has_d=true, .has_w=true,
.has_reg=true, .has_mod=true, .has_rm=true},
// Immediate to register/memory // Immediate to register/memory
{.id=2, .name="mov", .inst_enc=0b11000110, .mask_inst=0x1, .mask_w=0x1, .has_operands=true, {.id=2, .name="mov", .inst_enc=0b11000110, .mask_inst=0x1, .mask_w=0x1, .has_operands=true,
.has_displacement=true, .has_data=true, .has_w=true, .has_mod=true, .has_rm=true}, .has_displacement=true, .has_data=true, .has_w=true, .has_mod=true, .has_rm=true},
// Immediate to register // Immediate to register
{.id=3, .name="mov", .inst_enc=0b10110000, .mask_inst=0xF, .mask_w=0x8, {.id=3, .name="mov", .inst_enc=0b10110000, .mask_inst=0xF, .mask_w=0x8,
.mask_reg=0x0b00000111, .has_data=true, .has_w=true}, .mask_reg=0b00000111, .has_data=true, .has_w=true},
// Memory to accumulator | Accumulator to memory using the `d` bit // Memory to accumulator | Accumulator to memory using the `d` bit
// even though the manual doesn't specify it // even though the manual doesn't specify it
{.id=4, .name="mov", .inst_enc=0b10100000, .mask_inst=0x3, .mask_w=0x1, {.id=4, .name="mov", .inst_enc=0b10100000, .mask_inst=0x3, .mask_w=0x1,
@ -153,35 +162,97 @@ InstFormat inst_formats[] =
Instruction parse_instruction(u8* buf) Instruction parse_instruction(u8* buf)
{ {
u8 inst = buf[0]; u8 inst = buf[0];
InstFormat format; InstFormat fmt;
bool matched_inst = false; bool matched_inst = false;
// TODO: This might be a good time to learn how to make a hashtable in C // TODO: This might be a good time to learn how to make a hashtable in C
for (u16 i = 0; i < sizeof(inst_formats) / sizeof(InstFormat); i++) for (u16 i = 0; i < sizeof(inst_formats) / sizeof(InstFormat); i++)
{ {
if ((inst & ~inst_formats[i].mask_inst) == inst_formats[i].inst_enc) if ((inst & ~inst_formats[i].mask_inst) == inst_formats[i].inst_enc)
{ {
format = inst_formats[i]; fmt = inst_formats[i];
matched_inst = true; matched_inst = true;
break; break;
} }
} }
if (!matched_inst) if (!matched_inst)
return (Instruction){.bytes_read = 0}; return (Instruction){.bytes_read = 0};
u8_opt d_opt = none_u8();
u8_opt s_opt = none_u8();
u8_opt w_opt = none_u8();
u8_opt reg_opt = none_u8();
u8_opt mod_opt = none_u8();
u8_opt rm_opt = none_u8();
u16_opt data_opt = none_u16();
u16_opt displacement_opt = none_u16();
if (fmt.has_d) d_opt = some_u8((inst & 0x2) >> 1);
if (fmt.has_s) s_opt = some_u8((inst & 0x2) >> 1);
if (fmt.has_rm) rm_opt = some_u8(buf[1] & 0x7);
if (fmt.has_mod) mod_opt = some_u8((buf[1] & 0b11000000) >> 6);
if (fmt.has_w) w_opt = some_u8(mask_and_shift(buf[0], fmt.mask_w));
if (fmt.has_reg)
{
u8 reg = fmt.has_operands ? buf[1] : buf[0];
reg_opt = some_u8(mask_and_shift(reg, fmt.mask_reg));
}
if (fmt.has_data)
{
u8 idx = 1;
if (fmt.has_operands) idx += 1;
if (fmt.has_displacement) idx += mod_opt.value % 3;
u16 data = w_opt.value != 0 ? (i16)buf[idx+1] << 8 | buf[0] : (sbyte)buf[idx];
data_opt = some_u16(data);
}
if (fmt.has_displacement && mod_opt.value % 3 > 0)
{
u16 disp = mod_opt.value == MODE_MEM_DIS_16
? (i16)buf[3] << 8 | buf[2]
: (sbyte)buf[3];
displacement_opt = some_u16(disp);
}
u16 bytes_read = 1;
bytes_read += fmt.has_operands ? 1 : 0;
// This is a trick because mod == 1 and mod == 2 will displace one and two bytes
// respectively but mod == 3 wraps to 0 since it doesn't displace
if (fmt.has_displacement) bytes_read += mod_opt.value % 3;
if (fmt.has_data) bytes_read += w_opt.value == 0 ? 1 : 2;
return (Instruction) { return (Instruction) {
.id = format.id, .id = fmt.id,
.name = format.name, .name = fmt.name,
.data = none_u16(), .data = data_opt,
.displacement = none_u16(), .displacement = displacement_opt,
.w = none_u8(), .w = w_opt,
.d = none_u8(), .d = d_opt,
.s = none_u8(), .s = s_opt,
.mod = none_u8(), .mod = mod_opt,
.reg = none_u8(), .reg = reg_opt,
.rm = none_u8(), .rm = rm_opt,
.bytes_read = 2, .bytes_read = bytes_read,
}; };
} }
void decode_instruction(char* str_buf, Instruction inst)
{
IF_LET_SOME(u8, mod, inst.mod)
{
if (mod == MODE_RGSTR_MODE)
{
Register reg = registers[(size_t)inst.reg.value];
Register rm = registers[(size_t)inst.rm.value];
Register src_reg = inst.d.value == 0 ? reg : rm;
Register dst_reg = inst.d.value == 0 ? rm : reg;
char *src_name = reg_name(src_reg, inst.w.value);
char *dst_name = reg_name(dst_reg, inst.w.value);
sprintf(str_buf, "%s %s, %s ;%d", inst.name, dst_name, src_name, inst.id);
}
}
else
{
sprintf(str_buf, "%s ;%d", inst.name, inst.id);
}
}
bool mov_inst(FILE* f, unsigned char* buf, char inst) bool mov_inst(FILE* f, unsigned char* buf, char inst)
{ {
size_t bytes_read; size_t bytes_read;
@ -205,8 +276,6 @@ bool mov_inst(FILE* f, unsigned char* buf, char inst)
else else
{ {
bool is_direct_addr = mod == 0 && rm == 0b110; bool is_direct_addr = mod == 0 && rm == 0b110;
// This is a trick because mod == 1 and mod == 2 will displace one and two bytes
// respectively but mod == 3 wraps to 0 since it doesn't displace
int bytes_to_read = is_direct_addr ? 2 : mod % 3; int bytes_to_read = is_direct_addr ? 2 : mod % 3;
bytes_read = fread(buf, sizeof(char), bytes_to_read, f); bytes_read = fread(buf, sizeof(char), bytes_to_read, f);
char* eac_name = is_direct_addr ? "" : get_eac_registers(rm); char* eac_name = is_direct_addr ? "" : get_eac_registers(rm);
@ -381,21 +450,29 @@ int main(int argc, char** argv)
printf("; Decoded 8086 Assembly Instructions\n\n"); printf("; Decoded 8086 Assembly Instructions\n\n");
printf("bits 16\n\n"); printf("bits 16\n\n");
char *inst_str_buf = malloc(sizeof(char) * 256);
u32 bytes_processed = 0; u32 bytes_processed = 0;
while (bytes_processed < bytes_read) while (bytes_processed < bytes_read)
{ {
Instruction inst = parse_instruction(buffer+bytes_processed); Instruction inst = parse_instruction(buffer + bytes_processed);
bytes_processed += inst.bytes_read;
// char inst = buffer[0]; // char inst = buffer[0];
// if (mov_inst(f, buffer, inst)) goto handled; // if (mov_inst(f, buffer, inst)) goto handled;
// if (add_inst(f, buffer, inst)) goto handled; // if (add_inst(f, buffer, inst)) goto handled;
if (inst.name != NULL) if (inst.bytes_read > 0)
printf("%s ;%d", inst.name, inst.id); {
decode_instruction(inst_str_buf, inst);
printf("%s", inst_str_buf);
bytes_processed += inst.bytes_read;
}
else else
{
bytes_processed += 1;
fprintf(stderr, "___Unrecognized Instruction___"); fprintf(stderr, "___Unrecognized Instruction___");
}
// handled: // handled:
printf("\n"); printf("\n");
} }
free(inst_str_buf);
free(buffer); free(buffer);
} }

6
lib.h
View File

@ -23,7 +23,7 @@ enum OptionTag {NONE, SOME};
enum OptionTag tag; \ enum OptionTag tag; \
union { \ union { \
char none; \ char none; \
type some; \ type value; \
}; \ }; \
} type##_opt; \ } type##_opt; \
\ \
@ -34,13 +34,13 @@ enum OptionTag {NONE, SOME};
\ \
static inline type##_opt some_##type(type value) \ static inline type##_opt some_##type(type value) \
{ \ { \
return (type##_opt){ .tag = SOME, .some = value }; \ return (type##_opt){ .tag = SOME, .value = value }; \
} \ } \
\ \
static inline int get_some_##type(type##_opt opt, type* out_value) \ static inline int get_some_##type(type##_opt opt, type* out_value) \
{ \ { \
if (opt.tag != SOME) return 0; \ if (opt.tag != SOME) return 0; \
*out_value = opt.some; \ *out_value = opt.value; \
return 1; \ return 1; \
} }