#include "subsegs.h"
#include "dwarf2dbg.h"
#include "dw2gencfi.h"
-#include "opcode/i386.h"
#include "elf/x86-64.h"
#ifndef REGISTER_WARNINGS
CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
{".ssse3", PROCESSOR_UNKNOWN,
CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSSE3},
+ {".sse4.1", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSSE3|CpuSSE4_1},
{".3dnow", PROCESSOR_UNKNOWN,
CpuMMX|Cpu3dnow},
{".3dnowa", PROCESSOR_UNKNOWN,
if (prefix >= REX_OPCODE && prefix < REX_OPCODE + 16
&& flag_code == CODE_64BIT)
{
- if ((i.prefix[REX_PREFIX] & prefix & REX_MODE64)
- || ((i.prefix[REX_PREFIX] & (REX_EXTX | REX_EXTY | REX_EXTZ))
- && (prefix & (REX_EXTX | REX_EXTY | REX_EXTZ))))
+ if ((i.prefix[REX_PREFIX] & prefix & REX_W)
+ || ((i.prefix[REX_PREFIX] & (REX_R | REX_X | REX_B))
+ && (prefix & (REX_R | REX_X | REX_B))))
ret = 0;
q = REX_PREFIX;
}
reg_hash = hash_new ();
{
const reg_entry *regtab;
+ unsigned int regtab_size = i386_regtab_size;
- for (regtab = i386_regtab;
- regtab < i386_regtab + sizeof (i386_regtab) / sizeof (i386_regtab[0]);
- regtab++)
+ for (regtab = i386_regtab; regtab_size--; regtab++)
{
hash_err = hash_insert (reg_hash, regtab->reg_name, (PTR) regtab);
if (hash_err)
#endif
digit_chars['-'] = '-';
mnemonic_chars['-'] = '-';
+ mnemonic_chars['.'] = '.';
identifier_chars['_'] = '_';
identifier_chars['.'] = '.';
fprintf (stdout, " sib: base %x index %x scale %x\n",
x->sib.base, x->sib.index, x->sib.scale);
fprintf (stdout, " rex: 64bit %x extX %x extY %x extZ %x\n",
- (x->rex & REX_MODE64) != 0,
- (x->rex & REX_EXTX) != 0,
- (x->rex & REX_EXTY) != 0,
- (x->rex & REX_EXTZ) != 0);
+ (x->rex & REX_W) != 0,
+ (x->rex & REX_R) != 0,
+ (x->rex & REX_X) != 0,
+ (x->rex & REX_B) != 0);
for (i = 0; i < x->operands; i++)
{
fprintf (stdout, " #%d: ", i + 1);
/* Undo SYSV386_COMPAT brokenness when in Intel mode. See i386.h */
if (SYSV386_COMPAT
&& (i.tm.base_opcode & 0xfffffde0) == 0xdce0)
- i.tm.base_opcode ^= FloatR;
+ i.tm.base_opcode ^= Opcode_FloatR;
/* Zap movzx and movsx suffix. The suffix may have been set from
"word ptr" or "byte ptr" on the source operand, but we'll use
}
if ((i.tm.opcode_modifier & Rex64) != 0)
- i.rex |= REX_MODE64;
+ i.rex |= REX_W;
/* For 8 bit registers we need an empty rex prefix. Also if the
instruction already has a prefix, we need to convert old
continue;
break;
case 2:
+ /* xchg %eax, %eax is a special case. It is an aliase for nop
+ only in 32bit mode and we can use opcode 0x90. In 64bit
+ mode, we can't use 0x90 for xchg %eax, %eax since it should
+ zero-extend %eax to %rax. */
+ if (flag_code == CODE_64BIT
+ && t->base_opcode == 0x90
+ && i.types [0] == (Acc | Reg32)
+ && i.types [1] == (Acc | Reg32))
+ continue;
case 3:
case 4:
overlap1 = i.types[1] & operand_types[1];
}
/* found_reverse_match holds which of D or FloatDR
we've found. */
- found_reverse_match = t->opcode_modifier & (D | FloatDR);
+ if ((t->opcode_modifier & D))
+ found_reverse_match = Opcode_D;
+ else if ((t->opcode_modifier & FloatD))
+ found_reverse_match = Opcode_FloatD;
+ else
+ found_reverse_match = 0;
+ if ((t->opcode_modifier & FloatR))
+ found_reverse_match |= Opcode_FloatR;
}
else
{
if (i.operands != 2
|| i.types [0] != (Acc | Reg64)
|| i.types [1] != (Acc | Reg64)
- || strcmp (i.tm.name, "xchg") != 0)
- i.rex |= REX_MODE64;
+ || i.tm.base_opcode != 0x90)
+ i.rex |= REX_W;
}
/* Size floating point instruction. */
is converted into xor %reg, %reg. */
if (i.tm.opcode_modifier & regKludge)
{
- unsigned int first_reg_op = (i.types[0] & Reg) ? 0 : 1;
- /* Pretend we saw the extra register operand. */
- assert (i.reg_operands == 1
- && i.op[first_reg_op + 1].regs == 0);
- i.op[first_reg_op + 1].regs = i.op[first_reg_op].regs;
- i.types[first_reg_op + 1] = i.types[first_reg_op];
- i.operands++;
- i.reg_operands++;
+ if ((i.tm.cpu_flags & CpuSSE4_1))
+ {
+ /* The first operand in instruction blendvpd, blendvps and
+ pblendvb in SSE4.1 is implicit and must be xmm0. */
+ assert (i.operands == 3
+ && i.reg_operands >= 2
+ && i.types[0] == RegXMM);
+ if (i.op[0].regs->reg_num != 0)
+ {
+ if (intel_syntax)
+ as_bad (_("the last operand of `%s' must be `%sxmm0'"),
+ i.tm.name, register_prefix);
+ else
+ as_bad (_("the first operand of `%s' must be `%sxmm0'"),
+ i.tm.name, register_prefix);
+ return 0;
+ }
+ i.op[0] = i.op[1];
+ i.op[1] = i.op[2];
+ i.types[0] = i.types[1];
+ i.types[1] = i.types[2];
+ i.operands--;
+ i.reg_operands--;
+
+ /* We need to adjust fields in i.tm since they are used by
+ build_modrm_byte. */
+ i.tm.operand_types [0] = i.tm.operand_types [1];
+ i.tm.operand_types [1] = i.tm.operand_types [2];
+ i.tm.operands--;
+ }
+ else
+ {
+ unsigned int first_reg_op = (i.types[0] & Reg) ? 0 : 1;
+ /* Pretend we saw the extra register operand. */
+ assert (i.reg_operands == 1
+ && i.op[first_reg_op + 1].regs == 0);
+ i.op[first_reg_op + 1].regs = i.op[first_reg_op].regs;
+ i.types[first_reg_op + 1] = i.types[first_reg_op];
+ i.operands++;
+ i.reg_operands++;
+ }
}
if (i.tm.opcode_modifier & ShortForm)
}
i.tm.base_opcode |= (i.op[0].regs->reg_num << 3);
if ((i.op[0].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
}
else
{
/* Register goes in low 3 bits of opcode. */
i.tm.base_opcode |= i.op[op].regs->reg_num;
if ((i.op[op].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
if (!quiet_warnings && (i.tm.opcode_modifier & Ugh) != 0)
{
/* Warn about some common errors, but press on regardless.
default_seg = build_modrm_byte ();
}
- else if ((i.tm.base_opcode & ~(D | W)) == MOV_AX_DISP32)
+ else if ((i.tm.base_opcode & ~0x3) == MOV_AX_DISP32)
{
default_seg = &ds;
}
destination operand, then we assume the source operand may
sometimes be a memory operand and so we need to store the
destination in the i.rm.reg field. */
- if ((i.tm.operand_types[dest] & AnyMem) == 0)
+ if ((i.tm.operand_types[dest] & (AnyMem | RegMem)) == 0)
{
i.rm.reg = i.op[dest].regs->reg_num;
i.rm.regmem = i.op[source].regs->reg_num;
if ((i.op[dest].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTX;
+ i.rex |= REX_R;
if ((i.op[source].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
}
else
{
i.rm.reg = i.op[source].regs->reg_num;
i.rm.regmem = i.op[dest].regs->reg_num;
if ((i.op[dest].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
if ((i.op[source].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTX;
+ i.rex |= REX_R;
}
- if (flag_code != CODE_64BIT && (i.rex & (REX_EXTX | REX_EXTZ)))
+ if (flag_code != CODE_64BIT && (i.rex & (REX_R | REX_B)))
{
if (!((i.types[0] | i.types[1]) & Control))
abort ();
- i.rex &= ~(REX_EXTX | REX_EXTZ);
+ i.rex &= ~(REX_R | REX_B);
add_prefix (LOCK_PREFIX_OPCODE);
}
}
else
i.types[op] |= Disp32S;
if ((i.index_reg->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTY;
+ i.rex |= REX_X;
}
}
/* RIP addressing for 64bit mode. */
i.rm.regmem = i.base_reg->reg_num;
if ((i.base_reg->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
i.sib.base = i.base_reg->reg_num;
/* x86-64 ignores REX prefix bit here to avoid decoder
complications. */
i.sib.index = i.index_reg->reg_num;
i.rm.regmem = ESCAPE_TO_TWO_BYTE_ADDRESSING;
if ((i.index_reg->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTY;
+ i.rex |= REX_X;
}
if (i.disp_operands
{
i.rm.regmem = i.op[op].regs->reg_num;
if ((i.op[op].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTZ;
+ i.rex |= REX_B;
}
else
{
i.rm.reg = i.op[op].regs->reg_num;
if ((i.op[op].regs->reg_flags & RegRex) != 0)
- i.rex |= REX_EXTX;
+ i.rex |= REX_R;
}
/* Now, if no memory operand has set i.rm.mode = 0, 1, 2 we
unsigned char *q;
unsigned int prefix;
- /* All opcodes on i386 have either 1 or 2 bytes. Supplemental
- Streaming SIMD extensions 3 Instructions have 3 bytes. We may
- use one more higher byte to specify a prefix the instruction
- requires. */
- if ((i.tm.cpu_flags & CpuSSSE3) != 0)
+ /* All opcodes on i386 have either 1 or 2 bytes. SSSE3 and
+ SSE4.1 instructions have 3 bytes. We may use one more higher
+ byte to specify a prefix the instruction requires. */
+ if ((i.tm.cpu_flags & (CpuSSSE3 | CpuSSE4_1)) != 0)
{
if (i.tm.base_opcode & 0xff000000)
{
}
else
{
- if ((i.tm.cpu_flags & CpuSSSE3) != 0)
+ if ((i.tm.cpu_flags & (CpuSSSE3 | CpuSSE4_1)) != 0)
{
p = frag_more (3);
*p++ = (i.tm.base_opcode >> 16) & 0xff;
know (e->X_op == O_register);
know (e->X_add_number >= 0
- && (valueT) e->X_add_number < ARRAY_SIZE (i386_regtab));
+ && (valueT) e->X_add_number < i386_regtab_size);
r = i386_regtab + e->X_add_number;
*end_op = input_line_pointer;
}