/* tc-i386.c -- Assemble code for the Intel 80386
- Copyright (C) 1989-2018 Free Software Foundation, Inc.
+ Copyright (C) 1989-2019 Free Software Foundation, Inc.
This file is part of GAS, the GNU Assembler.
#include "elf/x86-64.h"
#include "opcodes/i386-init.h"
+#ifdef HAVE_LIMITS_H
+#include <limits.h>
+#else
+#ifdef HAVE_SYS_PARAM_H
+#include <sys/param.h>
+#endif
+#ifndef INT_MAX
+#define INT_MAX (int) (((unsigned) (-1)) >> 1)
+#endif
+#endif
+
#ifndef REGISTER_WARNINGS
#define REGISTER_WARNINGS 1
#endif
vex256
} avxscalar;
+/* Encode VEX WIG instructions with specific vex.w. */
+static enum
+ {
+ vexw0 = 0,
+ vexw1
+ } vexwig;
+
/* Encode scalar EVEX LIG instructions with specific vector length. */
static enum
{
}
/* Check reverse. */
- gas_assert (i.operands == 2);
+ gas_assert (i.operands >= 2 && i.operands <= 3);
- for (j = 0; j < 2; j++)
+ for (j = 0; j < i.operands; j++)
{
+ unsigned int given = i.operands - j - 1;
+
if (t->operand_types[j].bitfield.reg
- && !match_operand_size (t, j, !j))
+ && !match_operand_size (t, j, given))
goto mismatch;
if (t->operand_types[j].bitfield.regsimd
- && !match_simd_size (t, j, !j))
+ && !match_simd_size (t, j, given))
goto mismatch;
if (t->operand_types[j].bitfield.acc
- && (!match_operand_size (t, j, !j) || !match_simd_size (t, j, !j)))
+ && (!match_operand_size (t, j, given)
+ || !match_simd_size (t, j, given)))
goto mismatch;
- if ((i.flags[!j] & Operand_Mem) && !match_mem_size (t, j, !j))
+ if ((i.flags[given] & Operand_Mem) && !match_mem_size (t, j, given))
goto mismatch;
}
unsigned int register_specifier;
unsigned int implied_prefix;
unsigned int vector_length;
+ unsigned int w;
/* Check register specifier. */
if (i.vex.register_specifier)
else
register_specifier = 0xf;
- /* Use 2-byte VEX prefix by swapping destination and source
- operand. */
- if (i.vec_encoding != vex_encoding_vex3
+ /* Use 2-byte VEX prefix by swapping destination and source operand
+ if there are more than 1 register operand. */
+ if (i.reg_operands > 1
+ && i.vec_encoding != vex_encoding_vex3
&& i.dir_encoding == dir_encoding_default
&& i.operands == i.reg_operands
&& operand_type_equal (&i.types[0], &i.types[i.operands - 1])
abort ();
}
+ /* Check the REX.W bit and VEXW. */
+ if (i.tm.opcode_modifier.vexw == VEXWIG)
+ w = (vexwig == vexw1 || (i.rex & REX_W)) ? 1 : 0;
+ else if (i.tm.opcode_modifier.vexw)
+ w = i.tm.opcode_modifier.vexw == VEXW1 ? 1 : 0;
+ else
+ w = (flag_code == CODE_64BIT ? i.rex & REX_W : vexwig == vexw1) ? 1 : 0;
+
/* Use 2-byte VEX prefix if possible. */
- if (i.vec_encoding != vex_encoding_vex3
+ if (w == 0
+ && i.vec_encoding != vex_encoding_vex3
&& i.tm.opcode_modifier.vexopcode == VEX0F
- && i.tm.opcode_modifier.vexw != VEXW1
&& (i.rex & (REX_W | REX_X | REX_B)) == 0)
{
/* 2-byte VEX prefix. */
else
{
/* 3-byte VEX prefix. */
- unsigned int m, w;
+ unsigned int m;
i.vex.length = 3;
of RXB bits from REX. */
i.vex.bytes[1] = (~i.rex & 0x7) << 5 | m;
- /* Check the REX.W bit. */
- w = (i.rex & REX_W) ? 1 : 0;
- if (i.tm.opcode_modifier.vexw == VEXW1)
- w = 1;
-
i.vex.bytes[2] = (w << 7
| register_specifier << 3
| vector_length << 2
i.vrex &= ~vrex_used;
gas_assert (i.vrex == 0);
- /* Check the REX.W bit. */
- w = (i.rex & REX_W) ? 1 : 0;
- if (i.tm.opcode_modifier.vexw)
- {
- if (i.tm.opcode_modifier.vexw == VEXW1)
- w = 1;
- }
- /* If w is not set it means we are dealing with WIG instruction. */
- else if (!w)
- {
- if (evexwig == evexw1)
- w = 1;
- }
+ /* Check the REX.W bit and VEXW. */
+ if (i.tm.opcode_modifier.vexw == VEXWIG)
+ w = (evexwig == evexw1 || (i.rex & REX_W)) ? 1 : 0;
+ else if (i.tm.opcode_modifier.vexw)
+ w = i.tm.opcode_modifier.vexw == VEXW1 ? 1 : 0;
+ else
+ w = (flag_code == CODE_64BIT ? i.rex & REX_W : evexwig == evexw1) ? 1 : 0;
/* Encode the U bit. */
implied_prefix |= 0x4;
}
}
}
- else if (optimize > 1
- && i.reg_operands == 3
+ else if (i.reg_operands == 3
&& i.op[0].regs == i.op[1].regs
&& !i.types[2].bitfield.xmmword
&& (i.tm.opcode_modifier.vex
&& !i.rounding
&& is_evex_encoding (&i.tm)
&& (i.vec_encoding != vex_encoding_evex
+ || cpu_arch_isa_flags.bitfield.cpuavx512vl
|| i.tm.cpu_flags.bitfield.cpuavx512vl
|| (i.tm.operand_types[2].bitfield.zmmword
- && i.types[2].bitfield.ymmword)
- || cpu_arch_isa_flags.bitfield.cpuavx512vl)))
+ && i.types[2].bitfield.ymmword))))
&& ((i.tm.base_opcode == 0x55
|| i.tm.base_opcode == 0x6655
|| i.tm.base_opcode == 0x66df
|| i.tm.base_opcode == 0x6647)
&& i.tm.extension_opcode == None))
{
- /* Optimize: -O2:
+ /* Optimize: -O1:
VOP, one of vandnps, vandnpd, vxorps, vxorpd, vpsubb, vpsubd,
vpsubq and vpsubw:
EVEX VOP %zmmM, %zmmM, %zmmN
-> VEX VOP %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
EVEX VOP %ymmM, %ymmM, %ymmN
-> VEX VOP %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
VEX VOP %ymmM, %ymmM, %ymmN
-> VEX VOP %xmmM, %xmmM, %xmmN
VOP, one of vpandn and vpxor:
VOP, one of vpandnd and vpandnq:
EVEX VOP %zmmM, %zmmM, %zmmN
-> VEX vpandn %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
EVEX VOP %ymmM, %ymmM, %ymmN
-> VEX vpandn %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
VOP, one of vpxord and vpxorq:
EVEX VOP %zmmM, %zmmM, %zmmN
-> VEX vpxor %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
EVEX VOP %ymmM, %ymmM, %ymmN
-> VEX vpxor %xmmM, %xmmM, %xmmN (M and N < 16)
- -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16)
+ -> EVEX VOP %xmmM, %xmmM, %xmmN (M || N >= 16) (-O2)
VOP, one of kxord and kxorq:
VEX VOP %kM, %kM, %kN
-> VEX kxorw %kM, %kM, %kN
*/
if (is_evex_encoding (&i.tm))
{
- if (i.vec_encoding == vex_encoding_evex)
- i.tm.opcode_modifier.evex = EVEX128;
- else
+ if (i.vec_encoding != vex_encoding_evex)
{
i.tm.opcode_modifier.vex = VEX128;
i.tm.opcode_modifier.vexw = VEXW0;
i.tm.opcode_modifier.evex = 0;
}
+ else if (optimize > 1)
+ i.tm.opcode_modifier.evex = EVEX128;
+ else
+ return;
}
else if (i.tm.operand_types[0].bitfield.regmask)
{
i.types[j].bitfield.ymmword = 0;
}
}
+ else if ((cpu_arch_flags.bitfield.cpuavx
+ || cpu_arch_isa_flags.bitfield.cpuavx)
+ && i.vec_encoding != vex_encoding_evex
+ && !i.types[0].bitfield.zmmword
+ && !i.mask
+ && is_evex_encoding (&i.tm)
+ && (i.tm.base_opcode == 0x666f
+ || (i.tm.base_opcode ^ Opcode_SIMD_IntD) == 0x666f
+ || i.tm.base_opcode == 0xf36f
+ || (i.tm.base_opcode ^ Opcode_SIMD_IntD) == 0xf36f
+ || i.tm.base_opcode == 0xf26f
+ || (i.tm.base_opcode ^ Opcode_SIMD_IntD) == 0xf26f)
+ && i.tm.extension_opcode == None)
+ {
+ /* Optimize: -O1:
+ VOP, one of vmovdqa32, vmovdqa64, vmovdqu8, vmovdqu16,
+ vmovdqu32 and vmovdqu64:
+ EVEX VOP %xmmM, %xmmN
+ -> VEX vmovdqa|vmovdqu %xmmM, %xmmN (M and N < 16)
+ EVEX VOP %ymmM, %ymmN
+ -> VEX vmovdqa|vmovdqu %ymmM, %ymmN (M and N < 16)
+ EVEX VOP %xmmM, mem
+ -> VEX vmovdqa|vmovdqu %xmmM, mem (M < 16)
+ EVEX VOP %ymmM, mem
+ -> VEX vmovdqa|vmovdqu %ymmM, mem (M < 16)
+ EVEX VOP mem, %xmmN
+ -> VEX mvmovdqa|vmovdquem, %xmmN (N < 16)
+ EVEX VOP mem, %ymmN
+ -> VEX vmovdqa|vmovdqu mem, %ymmN (N < 16)
+ */
+ if (i.tm.base_opcode == 0xf26f)
+ i.tm.base_opcode = 0xf36f;
+ else if ((i.tm.base_opcode ^ Opcode_SIMD_IntD) == 0xf26f)
+ i.tm.base_opcode = 0xf36f ^ Opcode_SIMD_IntD;
+ i.tm.opcode_modifier.vex
+ = i.types[0].bitfield.ymmword ? VEX256 : VEX128;
+ i.tm.opcode_modifier.vexw = VEXW0;
+ i.tm.opcode_modifier.evex = 0;
+ i.tm.opcode_modifier.masking = 0;
+ i.tm.opcode_modifier.disp8memshift = 0;
+ i.memshift = 0;
+ for (j = 0; j < 2; j++)
+ if (operand_type_check (i.types[j], disp)
+ && i.op[j].disps->X_op == O_constant)
+ {
+ i.types[j].bitfield.disp8
+ = fits_in_disp8 (i.op[j].disps->X_add_number);
+ break;
+ }
+ }
}
/* This is the guts of the machine-dependent assembler. LINE points to a
}
/* If we are in 16-bit mode, do not allow addr16 or data16.
Similarly, in 32-bit mode, do not allow addr32 or data32. */
- if ((current_templates->start->opcode_modifier.size16
- || current_templates->start->opcode_modifier.size32)
+ if ((current_templates->start->opcode_modifier.size == SIZE16
+ || current_templates->start->opcode_modifier.size == SIZE32)
&& flag_code != CODE_64BIT
- && (current_templates->start->opcode_modifier.size32
+ && ((current_templates->start->opcode_modifier.size == SIZE32)
^ (flag_code == CODE_16BIT)))
{
as_bad (_("redundant %s prefix"),
if (!current_templates)
{
check_suffix:
- /* See if we can get a match by trimming off a suffix. */
- switch (mnem_p[-1])
+ if (mnem_p > mnemonic)
{
- case WORD_MNEM_SUFFIX:
- if (intel_syntax && (intel_float_operand (mnemonic) & 2))
- i.suffix = SHORT_MNEM_SUFFIX;
- else
- /* Fall through. */
- case BYTE_MNEM_SUFFIX:
- case QWORD_MNEM_SUFFIX:
- i.suffix = mnem_p[-1];
- mnem_p[-1] = '\0';
- current_templates = (const templates *) hash_find (op_hash,
- mnemonic);
- break;
- case SHORT_MNEM_SUFFIX:
- case LONG_MNEM_SUFFIX:
- if (!intel_syntax)
- {
- i.suffix = mnem_p[-1];
- mnem_p[-1] = '\0';
- current_templates = (const templates *) hash_find (op_hash,
- mnemonic);
- }
- break;
-
- /* Intel Syntax. */
- case 'd':
- if (intel_syntax)
+ /* See if we can get a match by trimming off a suffix. */
+ switch (mnem_p[-1])
{
- if (intel_float_operand (mnemonic) == 1)
+ case WORD_MNEM_SUFFIX:
+ if (intel_syntax && (intel_float_operand (mnemonic) & 2))
i.suffix = SHORT_MNEM_SUFFIX;
else
- i.suffix = LONG_MNEM_SUFFIX;
+ /* Fall through. */
+ case BYTE_MNEM_SUFFIX:
+ case QWORD_MNEM_SUFFIX:
+ i.suffix = mnem_p[-1];
mnem_p[-1] = '\0';
current_templates = (const templates *) hash_find (op_hash,
- mnemonic);
+ mnemonic);
+ break;
+ case SHORT_MNEM_SUFFIX:
+ case LONG_MNEM_SUFFIX:
+ if (!intel_syntax)
+ {
+ i.suffix = mnem_p[-1];
+ mnem_p[-1] = '\0';
+ current_templates = (const templates *) hash_find (op_hash,
+ mnemonic);
+ }
+ break;
+
+ /* Intel Syntax. */
+ case 'd':
+ if (intel_syntax)
+ {
+ if (intel_float_operand (mnemonic) == 1)
+ i.suffix = SHORT_MNEM_SUFFIX;
+ else
+ i.suffix = LONG_MNEM_SUFFIX;
+ mnem_p[-1] = '\0';
+ current_templates = (const templates *) hash_find (op_hash,
+ mnemonic);
+ }
+ break;
}
- break;
}
+
if (!current_templates)
{
as_bad (_("no such instruction: `%s'"), token_start);
&& i.types[0].bitfield.acc
&& operand_type_check (i.types[1], anymem))
continue;
+ /* Fall through. */
+
+ case 3:
if (!(size_match & MATCH_STRAIGHT))
goto check_reverse;
/* Reverse direction of operands if swapping is possible in the first
- the store form is requested, and the template is a load form,
- the non-default (swapped) form is requested. */
overlap1 = operand_type_and (operand_types[0], operand_types[1]);
- if (t->opcode_modifier.d && i.reg_operands == 2
+ if (t->opcode_modifier.d && i.reg_operands == i.operands
&& !operand_type_all_zero (&overlap1))
switch (i.dir_encoding)
{
case dir_encoding_default:
break;
}
- /* Fall through. */
-
- case 3:
/* If we want store form, we skip the current load. */
if ((i.dir_encoding == dir_encoding_store
|| i.dir_encoding == dir_encoding_swap)
if (!(size_match & MATCH_REVERSE))
continue;
/* Try reversing direction of operands. */
- overlap0 = operand_type_and (i.types[0], operand_types[1]);
- overlap1 = operand_type_and (i.types[1], operand_types[0]);
+ overlap0 = operand_type_and (i.types[0], operand_types[i.operands - 1]);
+ overlap1 = operand_type_and (i.types[i.operands - 1], operand_types[0]);
if (!operand_type_match (overlap0, i.types[0])
- || !operand_type_match (overlap1, i.types[1])
+ || !operand_type_match (overlap1, i.types[i.operands - 1])
|| (check_register
&& !operand_type_register_match (i.types[0],
- operand_types[1],
- i.types[1],
+ operand_types[i.operands - 1],
+ i.types[i.operands - 1],
operand_types[0])))
{
/* Does not match either direction. */
else if (operand_types[0].bitfield.tbyte)
found_reverse_match = Opcode_FloatD;
else if (operand_types[0].bitfield.xmmword
- || operand_types[1].bitfield.xmmword
+ || operand_types[i.operands - 1].bitfield.xmmword
|| operand_types[0].bitfield.regmmx
- || operand_types[1].bitfield.regmmx
+ || operand_types[i.operands - 1].bitfield.regmmx
|| is_any_vex_encoding(t))
found_reverse_match = (t->base_opcode & 0xee) != 0x6e
? Opcode_SIMD_FloatD : Opcode_SIMD_IntD;
i.tm.base_opcode ^= found_reverse_match;
- i.tm.operand_types[0] = operand_types[1];
- i.tm.operand_types[1] = operand_types[0];
+ i.tm.operand_types[0] = operand_types[i.operands - 1];
+ i.tm.operand_types[i.operands - 1] = operand_types[0];
}
return t;
{
/* If matched instruction specifies an explicit instruction mnemonic
suffix, use it. */
- if (i.tm.opcode_modifier.size16)
+ if (i.tm.opcode_modifier.size == SIZE16)
i.suffix = WORD_MNEM_SUFFIX;
- else if (i.tm.opcode_modifier.size32)
+ else if (i.tm.opcode_modifier.size == SIZE32)
i.suffix = LONG_MNEM_SUFFIX;
- else if (i.tm.opcode_modifier.size64)
+ else if (i.tm.opcode_modifier.size == SIZE64)
i.suffix = QWORD_MNEM_SUFFIX;
else if (i.reg_operands)
{
Destination register type is more significant than source
register type. crc32 in SSE4.2 prefers source register
type. */
- if (i.tm.base_opcode == 0xf20f38f1)
+ if (i.tm.base_opcode == 0xf20f38f0 && i.types[0].bitfield.reg)
{
- if (i.types[0].bitfield.reg && i.types[0].bitfield.word)
+ if (i.types[0].bitfield.byte)
+ i.suffix = BYTE_MNEM_SUFFIX;
+ else if (i.types[0].bitfield.word)
i.suffix = WORD_MNEM_SUFFIX;
- else if (i.types[0].bitfield.reg && i.types[0].bitfield.dword)
+ else if (i.types[0].bitfield.dword)
i.suffix = LONG_MNEM_SUFFIX;
- else if (i.types[0].bitfield.reg && i.types[0].bitfield.qword)
+ else if (i.types[0].bitfield.qword)
i.suffix = QWORD_MNEM_SUFFIX;
}
- else if (i.tm.base_opcode == 0xf20f38f0)
- {
- if (i.types[0].bitfield.reg && i.types[0].bitfield.byte)
- i.suffix = BYTE_MNEM_SUFFIX;
- }
if (!i.suffix)
{
int op;
- if (i.tm.base_opcode == 0xf20f38f1
- || i.tm.base_opcode == 0xf20f38f0)
+ if (i.tm.base_opcode == 0xf20f38f0)
{
/* We have to know the operand size for crc32. */
as_bad (_("ambiguous memory operand size for `%s`"),
if (!IS_ELF || !x86_used_note)
return;
- x86_isa_1_used |= GNU_PROPERTY_X86_UINT32_VALID;
x86_feature_2_used |= GNU_PROPERTY_X86_FEATURE_2_X86;
/* The .note.gnu.property section layout:
/* Check for "call/jmp *mem", "mov mem, %reg",
"test %reg, mem" and "binop mem, %reg" where binop
is one of adc, add, and, cmp, or, sbb, sub, xor
- instructions. Always generate R_386_GOT32X for
- "sym*GOT" operand in 32-bit mode. */
- if ((generate_relax_relocations
- || (!object_64bit
- && i.rm.mode == 0
- && i.rm.regmem == 5))
+ instructions without data prefix. Always generate
+ R_386_GOT32X for "sym*GOT" operand in 32-bit mode. */
+ if (i.prefix[DATA_PREFIX] == 0
+ && (generate_relax_relocations
+ || (!object_64bit
+ && i.rm.mode == 0
+ && i.rm.regmem == 5))
&& (i.rm.mode == 2
|| (i.rm.mode == 0 && i.rm.regmem == 5))
&& ((i.operands == 1
as_bad (_("missing or invalid expression `%s'"), save);
*input_line_pointer = c;
}
+ else if ((got_reloc == BFD_RELOC_386_PLT32
+ || got_reloc == BFD_RELOC_X86_64_PLT32)
+ && exp->X_op != O_symbol)
+ {
+ char c = *input_line_pointer;
+ *input_line_pointer = 0;
+ as_bad (_("invalid PLT expression `%s'"), save);
+ *input_line_pointer = c;
+ }
}
}
else
{
case BFD_RELOC_386_PLT32:
case BFD_RELOC_X86_64_PLT32:
- /* Make the jump instruction point to the address of the operand. At
- runtime we merely add the offset to the actual PLT entry. */
- value = -4;
+ /* Make the jump instruction point to the address of the operand.
+ At runtime we merely add the offset to the actual PLT entry.
+ NB: Subtract the offset size only for jump instructions. */
+ if (fixP->fx_pcrel)
+ value = -4;
break;
case BFD_RELOC_386_TLS_GD:
#define OPTION_MINTEL64 (OPTION_MD_BASE + 23)
#define OPTION_MFENCE_AS_LOCK_ADD (OPTION_MD_BASE + 24)
#define OPTION_X86_USED_NOTE (OPTION_MD_BASE + 25)
+#define OPTION_MVEXWIG (OPTION_MD_BASE + 26)
struct option md_longopts[] =
{
{"msse-check", required_argument, NULL, OPTION_MSSE_CHECK},
{"moperand-check", required_argument, NULL, OPTION_MOPERAND_CHECK},
{"mavxscalar", required_argument, NULL, OPTION_MAVXSCALAR},
+ {"mvexwig", required_argument, NULL, OPTION_MVEXWIG},
{"madd-bnd-prefix", no_argument, NULL, OPTION_MADD_BND_PREFIX},
{"mevexlig", required_argument, NULL, OPTION_MEVEXLIG},
{"mevexwig", required_argument, NULL, OPTION_MEVEXWIG},
as_fatal (_("invalid -mavxscalar= option: `%s'"), arg);
break;
+ case OPTION_MVEXWIG:
+ if (strcmp (arg, "0") == 0)
+ vexwig = evexw0;
+ else if (strcmp (arg, "1") == 0)
+ vexwig = evexw1;
+ else
+ as_fatal (_("invalid -mvexwig= option: `%s'"), arg);
+ break;
+
case OPTION_MADD_BND_PREFIX:
add_bnd_prefix = 1;
break;
{
optimize_for_space = 1;
/* Turn on all encoding optimizations. */
- optimize = -1;
+ optimize = INT_MAX;
}
else
{
encode scalar AVX instructions with specific vector\n\
length\n"));
fprintf (stream, _("\
+ -mvexwig=[0|1] (default: 0)\n\
+ encode VEX instructions with specific VEX.W value\n\
+ for VEX.W bit ignored instructions\n"));
+ fprintf (stream, _("\
-mevexlig=[128|256|512] (default: 128)\n\
encode scalar EVEX instructions with specific vector\n\
length\n"));
projects / deliverable / binutils-gdb.git / blobdiff