-/* i386.c -- Assemble code for the Intel 80386
+/* tc-i386.c -- Assemble code for the Intel 80386
Copyright 1989, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
- 2000, 2001, 2002, 2003, 2004, 2005, 2006
+ 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007
Free Software Foundation, Inc.
This file is part of GAS, the GNU Assembler.
#endif
#endif
-static INLINE unsigned int mode_from_disp_size PARAMS ((unsigned int));
-static INLINE int fits_in_signed_byte PARAMS ((offsetT));
-static INLINE int fits_in_unsigned_byte PARAMS ((offsetT));
-static INLINE int fits_in_unsigned_word PARAMS ((offsetT));
-static INLINE int fits_in_signed_word PARAMS ((offsetT));
-static INLINE int fits_in_unsigned_long PARAMS ((offsetT));
-static INLINE int fits_in_signed_long PARAMS ((offsetT));
-static int smallest_imm_type PARAMS ((offsetT));
-static offsetT offset_in_range PARAMS ((offsetT, int));
-static int add_prefix PARAMS ((unsigned int));
-static void set_code_flag PARAMS ((int));
-static void set_16bit_gcc_code_flag PARAMS ((int));
-static void set_intel_syntax PARAMS ((int));
-static void set_cpu_arch PARAMS ((int));
+static void set_code_flag (int);
+static void set_16bit_gcc_code_flag (int);
+static void set_intel_syntax (int);
+static void set_cpu_arch (int);
#ifdef TE_PE
-static void pe_directive_secrel PARAMS ((int));
+static void pe_directive_secrel (int);
#endif
-static void signed_cons PARAMS ((int));
-static char *output_invalid PARAMS ((int c));
-static int i386_operand PARAMS ((char *operand_string));
-static int i386_intel_operand PARAMS ((char *operand_string, int got_a_float));
-static const reg_entry *parse_register PARAMS ((char *reg_string,
- char **end_op));
-static char *parse_insn PARAMS ((char *, char *));
-static char *parse_operands PARAMS ((char *, const char *));
-static void swap_operands PARAMS ((void));
-static void optimize_imm PARAMS ((void));
-static void optimize_disp PARAMS ((void));
-static int match_template PARAMS ((void));
-static int check_string PARAMS ((void));
-static int process_suffix PARAMS ((void));
-static int check_byte_reg PARAMS ((void));
-static int check_long_reg PARAMS ((void));
-static int check_qword_reg PARAMS ((void));
-static int check_word_reg PARAMS ((void));
-static int finalize_imm PARAMS ((void));
-static int process_operands PARAMS ((void));
-static const seg_entry *build_modrm_byte PARAMS ((void));
-static void output_insn PARAMS ((void));
-static void output_branch PARAMS ((void));
-static void output_jump PARAMS ((void));
-static void output_interseg_jump PARAMS ((void));
-static void output_imm PARAMS ((fragS *insn_start_frag,
- offsetT insn_start_off));
-static void output_disp PARAMS ((fragS *insn_start_frag,
- offsetT insn_start_off));
+static void signed_cons (int);
+static char *output_invalid (int c);
+static int i386_operand (char *);
+static int i386_intel_operand (char *, int);
+static const reg_entry *parse_register (char *, char **);
+static char *parse_insn (char *, char *);
+static char *parse_operands (char *, const char *);
+static void swap_operands (void);
+static void swap_2_operands (int, int);
+static void optimize_imm (void);
+static void optimize_disp (void);
+static int match_template (void);
+static int check_string (void);
+static int process_suffix (void);
+static int check_byte_reg (void);
+static int check_long_reg (void);
+static int check_qword_reg (void);
+static int check_word_reg (void);
+static int finalize_imm (void);
+static int process_operands (void);
+static const seg_entry *build_modrm_byte (void);
+static void output_insn (void);
+static void output_imm (fragS *, offsetT);
+static void output_disp (fragS *, offsetT);
#ifndef I386COFF
-static void s_bss PARAMS ((int));
+static void s_bss (int);
#endif
#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)
static void handle_large_common (int small ATTRIBUTE_UNUSED);
/* Possible templates for current insn. */
static const templates *current_templates;
-/* Per instruction expressionS buffers: 2 displacements & 2 immediate max. */
-static expressionS disp_expressions[2], im_expressions[2];
+/* Per instruction expressionS buffers: max displacements & immediates. */
+static expressionS disp_expressions[MAX_MEMORY_OPERANDS];
+static expressionS im_expressions[MAX_IMMEDIATE_OPERANDS];
/* Current operand we are working on. */
static int this_operand;
/* 1 if register prefix % not required. */
static int allow_naked_reg = 0;
+/* Register prefix used for error message. */
+static const char *register_prefix = "%";
+
/* Used in 16 bit gcc mode to add an l suffix to call, ret, enter,
leave, push, and pop instructions so that gcc has the same stack
frame as in 32 bit mode. */
/* CPU feature flags. */
static unsigned int cpu_arch_flags = CpuUnknownFlags | CpuNo64;
+/* If we have selected a cpu we are generating instructions for. */
+static int cpu_arch_tune_set = 0;
+
+/* Cpu we are generating instructions for. */
+static enum processor_type cpu_arch_tune = PROCESSOR_UNKNOWN;
+
+/* CPU feature flags of cpu we are generating instructions for. */
+static unsigned int cpu_arch_tune_flags = 0;
+
+/* CPU instruction set architecture used. */
+static enum processor_type cpu_arch_isa = PROCESSOR_UNKNOWN;
+
+/* CPU feature flags of instruction set architecture used. */
+static unsigned int cpu_arch_isa_flags = 0;
+
/* If set, conditional jumps are not automatically promoted to handle
larger than a byte offset. */
static unsigned int no_cond_jump_promotion = 0;
{0, 0, 4, 0}
};
-static const arch_entry cpu_arch[] = {
- {"i8086", Cpu086 },
- {"i186", Cpu086|Cpu186 },
- {"i286", Cpu086|Cpu186|Cpu286 },
- {"i386", Cpu086|Cpu186|Cpu286|Cpu386 },
- {"i486", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486 },
- {"i586", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586 },
- {"i686", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686 },
- {"pentium", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586 },
- {"pentiumpro",Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686 },
- {"pentiumii", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX },
- {"pentiumiii",Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX|CpuMMX2|CpuSSE },
- {"pentium4", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX|CpuMMX2|CpuSSE|CpuSSE2 },
- {"prescott", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuPNI },
- {"k6", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuK6|CpuMMX },
- {"k6_2", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuK6|CpuMMX|Cpu3dnow },
- {"athlon", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6|CpuAthlon|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA },
- {"sledgehammer",Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6|CpuAthlon|CpuSledgehammer|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2 },
- {"opteron", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6|CpuAthlon|CpuSledgehammer|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2 },
- {".mmx", CpuMMX },
- {".sse", CpuMMX|CpuMMX2|CpuSSE },
- {".sse2", CpuMMX|CpuMMX2|CpuSSE|CpuSSE2 },
- {".sse3", CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3 },
- {".3dnow", CpuMMX|Cpu3dnow },
- {".3dnowa", CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA },
- {".padlock", CpuPadLock },
- {".pacifica", CpuSVME },
- {".svme", CpuSVME },
- {NULL, 0 }
+static const arch_entry cpu_arch[] =
+{
+ {"generic32", PROCESSOR_GENERIC32,
+ Cpu186|Cpu286|Cpu386},
+ {"generic64", PROCESSOR_GENERIC64,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2},
+ {"i8086", PROCESSOR_UNKNOWN,
+ 0},
+ {"i186", PROCESSOR_UNKNOWN,
+ Cpu186},
+ {"i286", PROCESSOR_UNKNOWN,
+ Cpu186|Cpu286},
+ {"i386", PROCESSOR_GENERIC32,
+ Cpu186|Cpu286|Cpu386},
+ {"i486", PROCESSOR_I486,
+ Cpu186|Cpu286|Cpu386|Cpu486},
+ {"i586", PROCESSOR_PENTIUM,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586},
+ {"i686", PROCESSOR_PENTIUMPRO,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686},
+ {"pentium", PROCESSOR_PENTIUM,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586},
+ {"pentiumpro",PROCESSOR_PENTIUMPRO,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686},
+ {"pentiumii", PROCESSOR_PENTIUMPRO,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX},
+ {"pentiumiii",PROCESSOR_PENTIUMPRO,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX|CpuMMX2|CpuSSE},
+ {"pentium4", PROCESSOR_PENTIUM4,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2},
+ {"prescott", PROCESSOR_NOCONA,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
+ {"nocona", PROCESSOR_NOCONA,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
+ {"yonah", PROCESSOR_CORE,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
+ {"core", PROCESSOR_CORE,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
+ {"merom", PROCESSOR_CORE2,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSSE3},
+ {"core2", PROCESSOR_CORE2,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuP4|CpuMMX
+ |CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSSE3},
+ {"k6", PROCESSOR_K6,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuK6|CpuMMX},
+ {"k6_2", PROCESSOR_K6,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuK6|CpuMMX|Cpu3dnow},
+ {"athlon", PROCESSOR_ATHLON,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6
+ |CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA},
+ {"sledgehammer", PROCESSOR_K8,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6
+ |CpuSledgehammer|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2},
+ {"opteron", PROCESSOR_K8,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6
+ |CpuSledgehammer|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2},
+ {"k8", PROCESSOR_K8,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6
+ |CpuSledgehammer|CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2},
+ {"amdfam10", PROCESSOR_AMDFAM10,
+ Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuK6|CpuSledgehammer
+ |CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA|CpuSSE|CpuSSE2|CpuSSE3|CpuSSE4a
+ |CpuABM},
+ {".mmx", PROCESSOR_UNKNOWN,
+ CpuMMX},
+ {".sse", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE},
+ {".sse2", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE|CpuSSE2},
+ {".sse3", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3},
+ {".ssse3", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSSE3},
+ {".3dnow", PROCESSOR_UNKNOWN,
+ CpuMMX|Cpu3dnow},
+ {".3dnowa", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|Cpu3dnow|Cpu3dnowA},
+ {".padlock", PROCESSOR_UNKNOWN,
+ CpuPadLock},
+ {".pacifica", PROCESSOR_UNKNOWN,
+ CpuSVME},
+ {".svme", PROCESSOR_UNKNOWN,
+ CpuSVME},
+ {".sse4a", PROCESSOR_UNKNOWN,
+ CpuMMX|CpuMMX2|CpuSSE|CpuSSE2|CpuSSE3|CpuSSE4a},
+ {".abm", PROCESSOR_UNKNOWN,
+ CpuABM}
};
const pseudo_typeS md_pseudo_table[] =
#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)
{"largecomm", handle_large_common, 0},
#else
- {"file", (void (*) PARAMS ((int))) dwarf2_directive_file, 0},
+ {"file", (void (*) (int)) dwarf2_directive_file, 0},
{"loc", dwarf2_directive_loc, 0},
{"loc_mark_labels", dwarf2_directive_loc_mark_labels, 0},
#endif
static struct hash_control *reg_hash;
\f
void
-i386_align_code (fragP, count)
- fragS *fragP;
- int count;
+i386_align_code (fragS *fragP, int count)
{
/* Various efficient no-op patterns for aligning code labels.
Note: Don't try to assemble the instructions in the comments.
static const char f32_1[] =
{0x90}; /* nop */
static const char f32_2[] =
- {0x89,0xf6}; /* movl %esi,%esi */
+ {0x66,0x90}; /* xchg %ax,%ax */
static const char f32_3[] =
{0x8d,0x76,0x00}; /* leal 0(%esi),%esi */
static const char f32_4[] =
f32_1, f32_2, f16_3, f16_4, f16_5, f16_6, f16_7, f16_8,
f32_15, f32_15, f32_15, f32_15, f32_15, f32_15, f32_15
};
+ /* nopl (%[re]ax) */
+ static const char alt_3[] =
+ {0x0f,0x1f,0x00};
+ /* nopl 0(%[re]ax) */
+ static const char alt_4[] =
+ {0x0f,0x1f,0x40,0x00};
+ /* nopl 0(%[re]ax,%[re]ax,1) */
+ static const char alt_5[] =
+ {0x0f,0x1f,0x44,0x00,0x00};
+ /* nopw 0(%[re]ax,%[re]ax,1) */
+ static const char alt_6[] =
+ {0x66,0x0f,0x1f,0x44,0x00,0x00};
+ /* nopl 0L(%[re]ax) */
+ static const char alt_7[] =
+ {0x0f,0x1f,0x80,0x00,0x00,0x00,0x00};
+ /* nopl 0L(%[re]ax,%[re]ax,1) */
+ static const char alt_8[] =
+ {0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* nopw 0L(%[re]ax,%[re]ax,1) */
+ static const char alt_9[] =
+ {0x66,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_10[] =
+ {0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* data16
+ nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_long_11[] =
+ {0x66,
+ 0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* data16
+ data16
+ nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_long_12[] =
+ {0x66,
+ 0x66,
+ 0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* data16
+ data16
+ data16
+ nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_long_13[] =
+ {0x66,
+ 0x66,
+ 0x66,
+ 0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* data16
+ data16
+ data16
+ data16
+ nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_long_14[] =
+ {0x66,
+ 0x66,
+ 0x66,
+ 0x66,
+ 0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* data16
+ data16
+ data16
+ data16
+ data16
+ nopw %cs:0L(%[re]ax,%[re]ax,1) */
+ static const char alt_long_15[] =
+ {0x66,
+ 0x66,
+ 0x66,
+ 0x66,
+ 0x66,
+ 0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ /* nopl 0(%[re]ax,%[re]ax,1)
+ nopw 0(%[re]ax,%[re]ax,1) */
+ static const char alt_short_11[] =
+ {0x0f,0x1f,0x44,0x00,0x00,
+ 0x66,0x0f,0x1f,0x44,0x00,0x00};
+ /* nopw 0(%[re]ax,%[re]ax,1)
+ nopw 0(%[re]ax,%[re]ax,1) */
+ static const char alt_short_12[] =
+ {0x66,0x0f,0x1f,0x44,0x00,0x00,
+ 0x66,0x0f,0x1f,0x44,0x00,0x00};
+ /* nopw 0(%[re]ax,%[re]ax,1)
+ nopl 0L(%[re]ax) */
+ static const char alt_short_13[] =
+ {0x66,0x0f,0x1f,0x44,0x00,0x00,
+ 0x0f,0x1f,0x80,0x00,0x00,0x00,0x00};
+ /* nopl 0L(%[re]ax)
+ nopl 0L(%[re]ax) */
+ static const char alt_short_14[] =
+ {0x0f,0x1f,0x80,0x00,0x00,0x00,0x00,
+ 0x0f,0x1f,0x80,0x00,0x00,0x00,0x00};
+ /* nopl 0L(%[re]ax)
+ nopl 0L(%[re]ax,%[re]ax,1) */
+ static const char alt_short_15[] =
+ {0x0f,0x1f,0x80,0x00,0x00,0x00,0x00,
+ 0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00};
+ static const char *const alt_short_patt[] = {
+ f32_1, f32_2, alt_3, alt_4, alt_5, alt_6, alt_7, alt_8,
+ alt_9, alt_10, alt_short_11, alt_short_12, alt_short_13,
+ alt_short_14, alt_short_15
+ };
+ static const char *const alt_long_patt[] = {
+ f32_1, f32_2, alt_3, alt_4, alt_5, alt_6, alt_7, alt_8,
+ alt_9, alt_10, alt_long_11, alt_long_12, alt_long_13,
+ alt_long_14, alt_long_15
+ };
if (count <= 0 || count > 15)
return;
- /* The recommended way to pad 64bit code is to use NOPs preceded by
- maximally four 0x66 prefixes. Balance the size of nops. */
- if (flag_code == CODE_64BIT)
+ /* We need to decide which NOP sequence to use for 32bit and
+ 64bit. When -mtune= is used:
+
+ 1. For PROCESSOR_I486, PROCESSOR_PENTIUM and PROCESSOR_GENERIC32,
+ f32_patt will be used.
+ 2. For PROCESSOR_K8 and PROCESSOR_AMDFAM10 in 64bit, NOPs with 0x66 prefix will be used.
+ 3. For PROCESSOR_CORE2, alt_long_patt will be used.
+ 4. For PROCESSOR_PENTIUMPRO, PROCESSOR_PENTIUM4, PROCESSOR_NOCONA,
+ PROCESSOR_CORE, PROCESSOR_CORE2, PROCESSOR_K6, PROCESSOR_ATHLON
+ and PROCESSOR_GENERIC64, alt_short_patt will be used.
+
+ When -mtune= isn't used, alt_short_patt will be used if
+ cpu_arch_isa_flags has Cpu686. Otherwise, f32_patt will be used.
+
+ When -march= or .arch is used, we can't use anything beyond
+ cpu_arch_isa_flags. */
+
+ if (flag_code == CODE_16BIT)
+ {
+ memcpy (fragP->fr_literal + fragP->fr_fix,
+ f16_patt[count - 1], count);
+ if (count > 8)
+ /* Adjust jump offset. */
+ fragP->fr_literal[fragP->fr_fix + 1] = count - 2;
+ }
+ else if (flag_code == CODE_64BIT && cpu_arch_tune == PROCESSOR_K8)
{
int i;
int nnops = (count + 3) / 4;
int remains = count - nnops * len;
int pos = 0;
+ /* The recommended way to pad 64bit code is to use NOPs preceded
+ by maximally four 0x66 prefixes. Balance the size of nops. */
for (i = 0; i < remains; i++)
{
memset (fragP->fr_literal + fragP->fr_fix + pos, 0x66, len);
}
}
else
- if (flag_code == CODE_16BIT)
- {
- memcpy (fragP->fr_literal + fragP->fr_fix,
- f16_patt[count - 1], count);
- if (count > 8)
- /* Adjust jump offset. */
- fragP->fr_literal[fragP->fr_fix + 1] = count - 2;
- }
- else
+ {
+ const char *const *patt = NULL;
+
+ if (cpu_arch_isa == PROCESSOR_UNKNOWN)
+ {
+ /* PROCESSOR_UNKNOWN means that all ISAs may be used. */
+ switch (cpu_arch_tune)
+ {
+ case PROCESSOR_UNKNOWN:
+ /* We use cpu_arch_isa_flags to check if we SHOULD
+ optimize for Cpu686. */
+ if ((cpu_arch_isa_flags & Cpu686) != 0)
+ patt = alt_short_patt;
+ else
+ patt = f32_patt;
+ break;
+ case PROCESSOR_CORE2:
+ patt = alt_long_patt;
+ break;
+ case PROCESSOR_PENTIUMPRO:
+ case PROCESSOR_PENTIUM4:
+ case PROCESSOR_NOCONA:
+ case PROCESSOR_CORE:
+ case PROCESSOR_K6:
+ case PROCESSOR_ATHLON:
+ case PROCESSOR_K8:
+ case PROCESSOR_GENERIC64:
+ case PROCESSOR_AMDFAM10:
+ patt = alt_short_patt;
+ break;
+ case PROCESSOR_I486:
+ case PROCESSOR_PENTIUM:
+ case PROCESSOR_GENERIC32:
+ patt = f32_patt;
+ break;
+ }
+ }
+ else
+ {
+ switch (cpu_arch_tune)
+ {
+ case PROCESSOR_UNKNOWN:
+ /* When cpu_arch_isa is net, cpu_arch_tune shouldn't be
+ PROCESSOR_UNKNOWN. */
+ abort ();
+ break;
+
+ case PROCESSOR_I486:
+ case PROCESSOR_PENTIUM:
+ case PROCESSOR_PENTIUMPRO:
+ case PROCESSOR_PENTIUM4:
+ case PROCESSOR_NOCONA:
+ case PROCESSOR_CORE:
+ case PROCESSOR_K6:
+ case PROCESSOR_ATHLON:
+ case PROCESSOR_K8:
+ case PROCESSOR_AMDFAM10:
+ case PROCESSOR_GENERIC32:
+ /* We use cpu_arch_isa_flags to check if we CAN optimize
+ for Cpu686. */
+ if ((cpu_arch_isa_flags & Cpu686) != 0)
+ patt = alt_short_patt;
+ else
+ patt = f32_patt;
+ break;
+ case PROCESSOR_CORE2:
+ if ((cpu_arch_isa_flags & Cpu686) != 0)
+ patt = alt_long_patt;
+ else
+ patt = f32_patt;
+ break;
+ case PROCESSOR_GENERIC64:
+ patt = alt_short_patt;
+ break;
+ }
+ }
+
memcpy (fragP->fr_literal + fragP->fr_fix,
- f32_patt[count - 1], count);
+ patt[count - 1], count);
+ }
fragP->fr_var = count;
}
static INLINE unsigned int
-mode_from_disp_size (t)
- unsigned int t;
+mode_from_disp_size (unsigned int t)
{
return (t & Disp8) ? 1 : (t & (Disp16 | Disp32 | Disp32S)) ? 2 : 0;
}
static INLINE int
-fits_in_signed_byte (num)
- offsetT num;
+fits_in_signed_byte (offsetT num)
{
return (num >= -128) && (num <= 127);
}
static INLINE int
-fits_in_unsigned_byte (num)
- offsetT num;
+fits_in_unsigned_byte (offsetT num)
{
return (num & 0xff) == num;
}
static INLINE int
-fits_in_unsigned_word (num)
- offsetT num;
+fits_in_unsigned_word (offsetT num)
{
return (num & 0xffff) == num;
}
static INLINE int
-fits_in_signed_word (num)
- offsetT num;
+fits_in_signed_word (offsetT num)
{
return (-32768 <= num) && (num <= 32767);
}
+
static INLINE int
-fits_in_signed_long (num)
- offsetT num ATTRIBUTE_UNUSED;
+fits_in_signed_long (offsetT num ATTRIBUTE_UNUSED)
{
#ifndef BFD64
return 1;
|| (((offsetT) -1 << 31) & num) == ((offsetT) -1 << 31));
#endif
} /* fits_in_signed_long() */
+
static INLINE int
-fits_in_unsigned_long (num)
- offsetT num ATTRIBUTE_UNUSED;
+fits_in_unsigned_long (offsetT num ATTRIBUTE_UNUSED)
{
#ifndef BFD64
return 1;
#endif
} /* fits_in_unsigned_long() */
-static int
-smallest_imm_type (num)
- offsetT num;
+static unsigned int
+smallest_imm_type (offsetT num)
{
- if (cpu_arch_flags != (Cpu086 | Cpu186 | Cpu286 | Cpu386 | Cpu486 | CpuNo64))
+ if (cpu_arch_flags != (Cpu186 | Cpu286 | Cpu386 | Cpu486 | CpuNo64))
{
/* This code is disabled on the 486 because all the Imm1 forms
in the opcode table are slower on the i486. They're the
}
static offsetT
-offset_in_range (val, size)
- offsetT val;
- int size;
+offset_in_range (offsetT val, int size)
{
addressT mask;
class already exists, 1 if non rep/repne added, 2 if rep/repne
added. */
static int
-add_prefix (prefix)
- unsigned int prefix;
+add_prefix (unsigned int prefix)
{
int ret = 1;
unsigned int q;
}
static void
-set_code_flag (value)
- int value;
+set_code_flag (int value)
{
flag_code = value;
cpu_arch_flags &= ~(Cpu64 | CpuNo64);
}
static void
-set_16bit_gcc_code_flag (new_code_flag)
- int new_code_flag;
+set_16bit_gcc_code_flag (int new_code_flag)
{
flag_code = new_code_flag;
cpu_arch_flags &= ~(Cpu64 | CpuNo64);
}
static void
-set_intel_syntax (syntax_flag)
- int syntax_flag;
+set_intel_syntax (int syntax_flag)
{
/* Find out if register prefixing is specified. */
int ask_naked_reg = 0;
identifier_chars['%'] = intel_syntax && allow_naked_reg ? '%' : 0;
identifier_chars['$'] = intel_syntax ? '$' : 0;
+ register_prefix = allow_naked_reg ? "" : "%";
}
static void
-set_cpu_arch (dummy)
- int dummy ATTRIBUTE_UNUSED;
+set_cpu_arch (int dummy ATTRIBUTE_UNUSED)
{
SKIP_WHITESPACE ();
{
char *string = input_line_pointer;
int e = get_symbol_end ();
- int i;
+ unsigned int i;
- for (i = 0; cpu_arch[i].name; i++)
+ for (i = 0; i < ARRAY_SIZE (cpu_arch); i++)
{
if (strcmp (string, cpu_arch[i].name) == 0)
{
cpu_sub_arch_name = NULL;
cpu_arch_flags = (cpu_arch[i].flags
| (flag_code == CODE_64BIT ? Cpu64 : CpuNo64));
+ cpu_arch_isa = cpu_arch[i].type;
+ cpu_arch_isa_flags = cpu_arch[i].flags;
+ if (!cpu_arch_tune_set)
+ {
+ cpu_arch_tune = cpu_arch_isa;
+ cpu_arch_tune_flags = cpu_arch_isa_flags;
+ }
break;
}
if ((cpu_arch_flags | cpu_arch[i].flags) != cpu_arch_flags)
return;
}
}
- if (!cpu_arch[i].name)
+ if (i >= ARRAY_SIZE (cpu_arch))
as_bad (_("no such architecture: `%s'"), string);
*input_line_pointer = e;
}
void
-i386_print_statistics (file)
- FILE *file;
+i386_print_statistics (FILE *file)
{
hash_print_statistics (file, "i386 opcode", op_hash);
hash_print_statistics (file, "i386 register", reg_hash);
#ifdef DEBUG386
/* Debugging routines for md_assemble. */
-static void pi PARAMS ((char *, i386_insn *));
-static void pte PARAMS ((template *));
-static void pt PARAMS ((unsigned int));
-static void pe PARAMS ((expressionS *));
-static void ps PARAMS ((symbolS *));
+static void pte (template *);
+static void pt (unsigned int);
+static void pe (expressionS *);
+static void ps (symbolS *);
static void
-pi (line, x)
- char *line;
- i386_insn *x;
+pi (char *line, i386_insn *x)
{
unsigned int i;
}
static void
-pte (t)
- template *t;
+pte (template *t)
{
unsigned int i;
fprintf (stdout, " %d operands ", t->operands);
}
static void
-pe (e)
- expressionS *e;
+pe (expressionS *e)
{
fprintf (stdout, " operation %d\n", e->X_op);
fprintf (stdout, " add_number %ld (%lx)\n",
}
static void
-ps (s)
- symbolS *s;
+ps (symbolS *s)
{
fprintf (stdout, "%s type %s%s",
S_GET_NAME (s),
some cases we force the original symbol to be used. */
int
-tc_i386_fix_adjustable (fixP)
- fixS *fixP ATTRIBUTE_UNUSED;
+tc_i386_fix_adjustable (fixS *fixP ATTRIBUTE_UNUSED)
{
#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)
if (!IS_ELF)
return 1;
}
-static int intel_float_operand PARAMS ((const char *mnemonic));
-
static int
-intel_float_operand (mnemonic)
- const char *mnemonic;
+intel_float_operand (const char *mnemonic)
{
/* Note that the value returned is meaningful only for opcodes with (memory)
operands, hence the code here is free to improperly handle opcodes that
if (line == NULL)
return;
+ /* The order of the immediates should be reversed
+ for 2 immediates extrq and insertq instructions */
+ if ((i.imm_operands == 2)
+ && ((strcmp (mnemonic, "extrq") == 0)
+ || (strcmp (mnemonic, "insertq") == 0)))
+ {
+ swap_2_operands (0, 1);
+ /* "extrq" and insertq" are the only two instructions whose operands
+ have to be reversed even though they have two immediate operands.
+ */
+ if (intel_syntax)
+ swap_operands ();
+ }
+
/* Now we've parsed the mnemonic into a set of templates, and have the
operands at hand. */
/* All intel opcodes have reversed operands except for "bound" and
"enter". We also don't reverse intersegment "jmp" and "call"
instructions with 2 immediate operands so that the immediate segment
- precedes the offset, as it does when in AT&T mode. "enter" and the
- intersegment "jmp" and "call" instructions are the only ones that
- have two immediate operands. */
- if (intel_syntax && i.operands > 1
+ precedes the offset, as it does when in AT&T mode. */
+ if (intel_syntax
+ && i.operands > 1
&& (strcmp (mnemonic, "bound") != 0)
&& (strcmp (mnemonic, "invlpga") != 0)
&& !((i.types[0] & Imm) && (i.types[1] & Imm)))
{
expressionS *exp;
- if ((i.tm.cpu_flags & CpuPNI) && i.operands > 0)
+ if ((i.tm.cpu_flags & CpuSSE3) && i.operands > 0)
{
- /* These Intel Prescott New Instructions have the fixed
+ /* Streaming SIMD extensions 3 Instructions have the fixed
operands with an opcode suffix which is coded in the same
place as an 8-bit immediate field would be. Here we check
those operands and remove them afterwards. */
}
static char *
-parse_insn (line, mnemonic)
- char *line;
- char *mnemonic;
+parse_insn (char *line, char *mnemonic)
{
char *l = line;
char *token_start = l;
}
static char *
-parse_operands (l, mnemonic)
- char *l;
- const char *mnemonic;
+parse_operands (char *l, const char *mnemonic)
{
char *token_start;
}
static void
-swap_operands ()
+swap_2_operands (int xchg1, int xchg2)
{
union i386_op temp_op;
unsigned int temp_type;
enum bfd_reloc_code_real temp_reloc;
- int xchg1 = 0;
- int xchg2 = 0;
-
- if (i.operands == 2)
- {
- xchg1 = 0;
- xchg2 = 1;
- }
- else if (i.operands == 3)
- {
- xchg1 = 0;
- xchg2 = 2;
- }
+
temp_type = i.types[xchg2];
i.types[xchg2] = i.types[xchg1];
i.types[xchg1] = temp_type;
temp_reloc = i.reloc[xchg2];
i.reloc[xchg2] = i.reloc[xchg1];
i.reloc[xchg1] = temp_reloc;
+}
+
+static void
+swap_operands (void)
+{
+ switch (i.operands)
+ {
+ case 4:
+ swap_2_operands (1, i.operands - 2);
+ case 3:
+ case 2:
+ swap_2_operands (0, i.operands - 1);
+ break;
+ default:
+ abort ();
+ }
if (i.mem_operands == 2)
{
/* Try to ensure constant immediates are represented in the smallest
opcode possible. */
static void
-optimize_imm ()
+optimize_imm (void)
{
char guess_suffix = 0;
int op;
/* Try to use the smallest displacement type too. */
static void
-optimize_disp ()
+optimize_disp (void)
{
int op;
}
static int
-match_template ()
+match_template (void)
{
/* Points to template once we've found it. */
const template *t;
- unsigned int overlap0, overlap1, overlap2;
+ unsigned int overlap0, overlap1, overlap2, overlap3;
unsigned int found_reverse_match;
int suffix_check;
+ unsigned int operand_types [MAX_OPERANDS];
+ int addr_prefix_disp;
+ unsigned int j;
+
+#if MAX_OPERANDS != 4
+# error "MAX_OPERANDS must be 4."
+#endif
#define MATCH(overlap, given, template) \
((overlap & ~JumpAbsolute) \
overlap0 = 0;
overlap1 = 0;
overlap2 = 0;
+ overlap3 = 0;
found_reverse_match = 0;
+ for (j = 0; j < MAX_OPERANDS; j++)
+ operand_types [j] = 0;
+ addr_prefix_disp = -1;
suffix_check = (i.suffix == BYTE_MNEM_SUFFIX
? No_bSuf
: (i.suffix == WORD_MNEM_SUFFIX
for (t = current_templates->start; t < current_templates->end; t++)
{
+ addr_prefix_disp = -1;
+
/* Must have right number of operands. */
if (i.operands != t->operands)
continue;
&& (t->opcode_modifier & IgnoreSize)))
continue;
+ for (j = 0; j < MAX_OPERANDS; j++)
+ operand_types [j] = t->operand_types [j];
+
/* In general, don't allow 64-bit operands in 32-bit mode. */
if (i.suffix == QWORD_MNEM_SUFFIX
&& flag_code != CODE_64BIT
? (!(t->opcode_modifier & IgnoreSize)
&& !intel_float_operand (t->name))
: intel_float_operand (t->name) != 2)
- && (!(t->operand_types[0] & (RegMMX | RegXMM))
- || !(t->operand_types[t->operands > 1] & (RegMMX | RegXMM)))
+ && (!(operand_types[0] & (RegMMX | RegXMM))
+ || !(operand_types[t->operands > 1] & (RegMMX | RegXMM)))
&& (t->base_opcode != 0x0fc7
|| t->extension_opcode != 1 /* cmpxchg8b */))
continue;
break;
}
- overlap0 = i.types[0] & t->operand_types[0];
+ /* Address size prefix will turn Disp64/Disp32/Disp16 operand
+ into Disp32/Disp16/Disp32 operand. */
+ if (i.prefix[ADDR_PREFIX] != 0)
+ {
+ unsigned int DispOn = 0, DispOff = 0;
+
+ switch (flag_code)
+ {
+ case CODE_16BIT:
+ DispOn = Disp32;
+ DispOff = Disp16;
+ break;
+ case CODE_32BIT:
+ DispOn = Disp16;
+ DispOff = Disp32;
+ break;
+ case CODE_64BIT:
+ DispOn = Disp32;
+ DispOff = Disp64;
+ break;
+ }
+
+ for (j = 0; j < MAX_OPERANDS; j++)
+ {
+ /* There should be only one Disp operand. */
+ if ((operand_types[j] & DispOff))
+ {
+ addr_prefix_disp = j;
+ operand_types[j] |= DispOn;
+ operand_types[j] &= ~DispOff;
+ break;
+ }
+ }
+ }
+
+ overlap0 = i.types[0] & operand_types[0];
switch (t->operands)
{
case 1:
- if (!MATCH (overlap0, i.types[0], t->operand_types[0]))
+ if (!MATCH (overlap0, i.types[0], operand_types[0]))
continue;
break;
case 2:
case 3:
- overlap1 = i.types[1] & t->operand_types[1];
- if (!MATCH (overlap0, i.types[0], t->operand_types[0])
- || !MATCH (overlap1, i.types[1], t->operand_types[1])
+ case 4:
+ overlap1 = i.types[1] & operand_types[1];
+ if (!MATCH (overlap0, i.types[0], operand_types[0])
+ || !MATCH (overlap1, i.types[1], operand_types[1])
/* monitor in SSE3 is a very special case. The first
- register and the second register may have differnet
+ register and the second register may have different
sizes. */
|| !((t->base_opcode == 0x0f01
&& t->extension_opcode == 0xc8)
|| CONSISTENT_REGISTER_MATCH (overlap0, i.types[0],
- t->operand_types[0],
+ operand_types[0],
overlap1, i.types[1],
- t->operand_types[1])))
+ operand_types[1])))
{
/* Check if other direction is valid ... */
if ((t->opcode_modifier & (D | FloatD)) == 0)
continue;
/* Try reversing direction of operands. */
- overlap0 = i.types[0] & t->operand_types[1];
- overlap1 = i.types[1] & t->operand_types[0];
- if (!MATCH (overlap0, i.types[0], t->operand_types[1])
- || !MATCH (overlap1, i.types[1], t->operand_types[0])
+ overlap0 = i.types[0] & operand_types[1];
+ overlap1 = i.types[1] & operand_types[0];
+ if (!MATCH (overlap0, i.types[0], operand_types[1])
+ || !MATCH (overlap1, i.types[1], operand_types[0])
|| !CONSISTENT_REGISTER_MATCH (overlap0, i.types[0],
- t->operand_types[1],
+ operand_types[1],
overlap1, i.types[1],
- t->operand_types[0]))
+ operand_types[0]))
{
/* Does not match either direction. */
continue;
we've found. */
found_reverse_match = t->opcode_modifier & (D | FloatDR);
}
- /* Found a forward 2 operand match here. */
- else if (t->operands == 3)
+ else
{
- /* Here we make use of the fact that there are no
- reverse match 3 operand instructions, and all 3
- operand instructions only need to be checked for
- register consistency between operands 2 and 3. */
- overlap2 = i.types[2] & t->operand_types[2];
- if (!MATCH (overlap2, i.types[2], t->operand_types[2])
- || !CONSISTENT_REGISTER_MATCH (overlap1, i.types[1],
- t->operand_types[1],
- overlap2, i.types[2],
- t->operand_types[2]))
+ /* Found a forward 2 operand match here. */
+ switch (t->operands)
+ {
+ case 4:
+ overlap3 = i.types[3] & operand_types[3];
+ case 3:
+ overlap2 = i.types[2] & operand_types[2];
+ break;
+ }
- continue;
+ switch (t->operands)
+ {
+ case 4:
+ if (!MATCH (overlap3, i.types[3], operand_types[3])
+ || !CONSISTENT_REGISTER_MATCH (overlap2,
+ i.types[2],
+ operand_types[2],
+ overlap3,
+ i.types[3],
+ operand_types[3]))
+ continue;
+ case 3:
+ /* Here we make use of the fact that there are no
+ reverse match 3 operand instructions, and all 3
+ operand instructions only need to be checked for
+ register consistency between operands 2 and 3. */
+ if (!MATCH (overlap2, i.types[2], operand_types[2])
+ || !CONSISTENT_REGISTER_MATCH (overlap1,
+ i.types[1],
+ operand_types[1],
+ overlap2,
+ i.types[2],
+ operand_types[2]))
+ continue;
+ break;
+ }
}
- /* Found either forward/reverse 2 or 3 operand match here:
+ /* Found either forward/reverse 2, 3 or 4 operand match here:
slip through to break. */
}
if (t->cpu_flags & ~cpu_arch_flags)
{
if (!intel_syntax
&& ((i.types[0] & JumpAbsolute)
- != (t->operand_types[0] & JumpAbsolute)))
+ != (operand_types[0] & JumpAbsolute)))
{
as_warn (_("indirect %s without `*'"), t->name);
}
/* Copy the template we found. */
i.tm = *t;
+
+ if (addr_prefix_disp != -1)
+ i.tm.operand_types[addr_prefix_disp]
+ = operand_types[addr_prefix_disp];
+
if (found_reverse_match)
{
/* If we found a reverse match we must alter the opcode
i.tm.base_opcode ^= found_reverse_match;
- i.tm.operand_types[0] = t->operand_types[1];
- i.tm.operand_types[1] = t->operand_types[0];
+ i.tm.operand_types[0] = operand_types[1];
+ i.tm.operand_types[1] = operand_types[0];
}
return 1;
}
static int
-check_string ()
+check_string (void)
{
int mem_op = (i.types[0] & AnyMem) ? 0 : 1;
if ((i.tm.operand_types[mem_op] & EsSeg) != 0)
if (i.suffix == QWORD_MNEM_SUFFIX
&& flag_code == CODE_64BIT
&& (i.tm.opcode_modifier & NoRex64) == 0)
- i.rex |= REX_MODE64;
+ {
+ /* Special case for xchg %rax,%rax. It is NOP and doesn't
+ need rex64. */
+ if (i.operands != 2
+ || i.types [0] != (Acc | Reg64)
+ || i.types [1] != (Acc | Reg64)
+ || strcmp (i.tm.name, "xchg") != 0)
+ i.rex |= REX_MODE64;
+ }
/* Size floating point instruction. */
if (i.suffix == LONG_MNEM_SUFFIX)
if (flag_code == CODE_64BIT
&& (i.tm.operand_types[op] & InOutPortReg) == 0)
{
- as_bad (_("Incorrect register `%%%s' used with `%c' suffix"),
- i.op[op].regs->reg_name,
+ as_bad (_("Incorrect register `%s%s' used with `%c' suffix"),
+ register_prefix, i.op[op].regs->reg_name,
i.suffix);
return 0;
}
}
static int
-check_long_reg ()
+check_long_reg (void)
{
int op;
lowering is more complicated. */
if (flag_code == CODE_64BIT)
{
- as_bad (_("Incorrect register `%%%s' used with `%c' suffix"),
- i.op[op].regs->reg_name,
+ as_bad (_("Incorrect register `%s%s' used with `%c' suffix"),
+ register_prefix, i.op[op].regs->reg_name,
i.suffix);
return 0;
}
else if ((i.types[op] & Reg64) != 0
&& (i.tm.operand_types[op] & (Reg32 | Acc)) != 0)
{
- as_bad (_("Incorrect register `%%%s' used with `%c' suffix"),
- i.op[op].regs->reg_name,
+ as_bad (_("Incorrect register `%s%s' used with `%c' suffix"),
+ register_prefix, i.op[op].regs->reg_name,
i.suffix);
return 0;
}
}
static int
-check_qword_reg ()
+check_qword_reg (void)
{
int op;
{
/* Prohibit these changes in the 64bit mode, since the
lowering is more complicated. */
- as_bad (_("Incorrect register `%%%s' used with `%c' suffix"),
- i.op[op].regs->reg_name,
+ as_bad (_("Incorrect register `%s%s' used with `%c' suffix"),
+ register_prefix, i.op[op].regs->reg_name,
i.suffix);
return 0;
}
}
static int
-check_word_reg ()
+check_word_reg (void)
{
int op;
for (op = i.operands; --op >= 0;)
lowering is more complicated. */
if (flag_code == CODE_64BIT)
{
- as_bad (_("Incorrect register `%%%s' used with `%c' suffix"),
- i.op[op].regs->reg_name,
+ as_bad (_("Incorrect register `%s%s' used with `%c' suffix"),
+ register_prefix, i.op[op].regs->reg_name,
i.suffix);
return 0;
}
}
static int
-finalize_imm ()
+finalize_imm (void)
{
unsigned int overlap0, overlap1, overlap2;
}
static int
-process_operands ()
+process_operands (void)
{
/* Default segment register this instruction will use for memory
accesses. 0 means unknown. This is only for optimizing out
{
unsigned int first_reg_op = (i.types[0] & Reg) ? 0 : 1;
/* Pretend we saw the extra register operand. */
- assert (i.op[first_reg_op + 1].regs == 0);
+ 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.reg_operands = 2;
+ i.operands++;
+ i.reg_operands++;
}
if (i.tm.opcode_modifier & ShortForm)
}
static const seg_entry *
-build_modrm_byte ()
+build_modrm_byte (void)
{
const seg_entry *default_seg = 0;
if (i.reg_operands == 2)
{
unsigned int source, dest;
- source = ((i.types[0]
- & (Reg | RegMMX | RegXMM
- | SReg2 | SReg3
- | Control | Debug | Test))
- ? 0 : 1);
+
+ switch (i.operands)
+ {
+ case 2:
+ source = 0;
+ break;
+ case 3:
+ /* When there are 3 operands, one of them may be immediate,
+ which may be the first or the last operand. Otherwise,
+ the first operand must be shift count register (cl). */
+ assert (i.imm_operands == 1
+ || (i.imm_operands == 0
+ && (i.types[0] & ShiftCount)));
+ source = (i.types[0] & (Imm | ShiftCount)) ? 1 : 0;
+ break;
+ case 4:
+ /* When there are 4 operands, the first two must be immediate
+ operands. The source operand will be the 3rd one. */
+ assert (i.imm_operands == 2
+ && (i.types[0] & Imm)
+ && (i.types[1] & Imm));
+ source = 2;
+ break;
+ default:
+ abort ();
+ }
+
dest = source + 1;
i.rm.mode = 3;
if (i.mem_operands)
{
unsigned int fake_zero_displacement = 0;
- unsigned int op = ((i.types[0] & AnyMem)
- ? 0
- : (i.types[1] & AnyMem) ? 1 : 2);
+ unsigned int op;
+
+ for (op = 0; op < i.operands; op++)
+ if ((i.types[op] & AnyMem))
+ break;
+ assert (op < i.operands);
default_seg = &ds;
i.rm.regmem = ESCAPE_TO_TWO_BYTE_ADDRESSING;
i.sib.base = NO_BASE_REGISTER;
i.sib.index = NO_INDEX_REGISTER;
- i.types[op] = ((i.prefix[ADDR_PREFIX] == 0) ? Disp32S : Disp32);
+ i.types[op] = ((i.prefix[ADDR_PREFIX] == 0)
+ ? Disp32S : Disp32);
}
- else if ((flag_code == CODE_16BIT) ^ (i.prefix[ADDR_PREFIX] != 0))
+ else if ((flag_code == CODE_16BIT)
+ ^ (i.prefix[ADDR_PREFIX] != 0))
{
i.rm.regmem = NO_BASE_REGISTER_16;
i.types[op] = Disp16;
i.rm.regmem = NO_BASE_REGISTER;
i.types[op] &= ~ Disp;
i.types[op] |= Disp32S;
- i.flags[op] = Operand_PCrel;
+ i.flags[op] |= Operand_PCrel;
if (! i.disp_operands)
fake_zero_displacement = 1;
}
{
if (flag_code == CODE_64BIT
&& (i.types[op] & Disp))
- i.types[op] = (i.types[op] & Disp8) | (i.prefix[ADDR_PREFIX] == 0 ? Disp32S : Disp32);
+ i.types[op] = ((i.types[op] & Disp8)
+ | (i.prefix[ADDR_PREFIX] == 0
+ ? Disp32S : Disp32));
i.rm.regmem = i.base_reg->reg_num;
if ((i.base_reg->reg_flags & RegRex) != 0)
registers are coded into the i.rm.reg field. */
if (i.reg_operands)
{
- unsigned int op =
- ((i.types[0]
- & (Reg | RegMMX | RegXMM
- | SReg2 | SReg3
- | Control | Debug | Test))
- ? 0
- : ((i.types[1]
- & (Reg | RegMMX | RegXMM
- | SReg2 | SReg3
- | Control | Debug | Test))
- ? 1
- : 2));
+ unsigned int op;
+
+ for (op = 0; op < i.operands; op++)
+ if ((i.types[op] & (Reg | RegMMX | RegXMM
+ | SReg2 | SReg3
+ | Control | Debug | Test)))
+ break;
+ assert (op < i.operands);
+
/* If there is an extension opcode to put here, the register
number must be put into the regmem field. */
if (i.tm.extension_opcode != None)
}
static void
-output_branch ()
+output_branch (void)
{
char *p;
int code16;
}
static void
-output_jump ()
+output_jump (void)
{
char *p;
int size;
}
static void
-output_interseg_jump ()
+output_interseg_jump (void)
{
char *p;
int size;
}
static void
-output_insn ()
+output_insn (void)
{
fragS *insn_start_frag;
offsetT insn_start_off;
unsigned char *q;
unsigned int prefix;
- /* All opcodes on i386 have either 1 or 2 bytes. Merom New
- Instructions have 3 bytes. We may use one more higher byte
- to specify a prefix the instruction requires. */
- if ((i.tm.cpu_flags & CpuMNI) != 0)
+ /* 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)
{
if (i.tm.base_opcode & 0xff000000)
{
}
else
{
- if ((i.tm.cpu_flags & CpuMNI) != 0)
+ if ((i.tm.cpu_flags & CpuSSSE3) != 0)
{
p = frag_more (3);
*p++ = (i.tm.base_opcode >> 16) & 0xff;
static int cons_sign = -1;
void
-x86_cons_fix_new (fragS *frag,
- unsigned int off,
- unsigned int len,
+x86_cons_fix_new (fragS *frag, unsigned int off, unsigned int len,
expressionS *exp)
{
enum bfd_reloc_code_real r = reloc (len, 0, cons_sign, got_reloc);
}
void
-x86_cons (exp, size)
- expressionS *exp;
- int size;
+x86_cons (expressionS *exp, int size)
{
if (size == 4 || (object_64bit && size == 8))
{
}
#endif
-static int i386_immediate PARAMS ((char *));
-
static int
-i386_immediate (imm_start)
- char *imm_start;
+i386_immediate (char *imm_start)
{
char *save_input_line_pointer;
char *gotfree_input_line;
if (i.imm_operands == MAX_IMMEDIATE_OPERANDS)
{
- as_bad (_("only 1 or 2 immediate operands are allowed"));
+ as_bad (_("at most %d immediate operands are allowed"),
+ MAX_IMMEDIATE_OPERANDS);
return 0;
}
return 0;
}
#endif
+ else if (!intel_syntax && exp->X_op == O_register)
+ {
+ as_bad (_("illegal immediate register operand %s"), imm_start);
+ return 0;
+ }
else
{
/* This is an address. The size of the address will be
return 1;
}
-static char *i386_scale PARAMS ((char *));
-
static char *
-i386_scale (scale)
- char *scale;
+i386_scale (char *scale)
{
offsetT val;
char *save = input_line_pointer;
return scale;
}
-static int i386_displacement PARAMS ((char *, char *));
-
static int
-i386_displacement (disp_start, disp_end)
- char *disp_start;
- char *disp_end;
+i386_displacement (char *disp_start, char *disp_end)
{
expressionS *exp;
segT exp_seg = 0;
int bigdisp, override;
unsigned int types = Disp;
+ if (i.disp_operands == MAX_MEMORY_OPERANDS)
+ {
+ as_bad (_("at most %d displacement operands are allowed"),
+ MAX_MEMORY_OPERANDS);
+ return 0;
+ }
+
if ((i.types[this_operand] & JumpAbsolute)
|| !(current_templates->start->opcode_modifier & (Jump | JumpDword)))
{
return 1;
}
-static int i386_index_check PARAMS ((const char *));
-
/* Make sure the memory operand we've been dealt is valid.
Return 1 on success, 0 on a failure. */
static int
-i386_index_check (operand_string)
- const char *operand_string;
+i386_index_check (const char *operand_string)
{
int ok;
#if INFER_ADDR_PREFIX
on error. */
static int
-i386_operand (operand_string)
- char *operand_string;
+i386_operand (char *operand_string)
{
const reg_entry *r;
char *end_op;
return 0;
}
\f
-static char output_invalid_buf[8];
+static char output_invalid_buf[sizeof (unsigned char) * 2 + 6];
static char *
-output_invalid (c)
- int c;
+output_invalid (int c)
{
if (ISPRINT (c))
- sprintf (output_invalid_buf, "'%c'", c);
+ snprintf (output_invalid_buf, sizeof (output_invalid_buf),
+ "'%c'", c);
else
- sprintf (output_invalid_buf, "(0x%x)", (unsigned) c);
+ snprintf (output_invalid_buf, sizeof (output_invalid_buf),
+ "(0x%x)", (unsigned char) c);
return output_invalid_buf;
}
#define OPTION_32 (OPTION_MD_BASE + 0)
#define OPTION_64 (OPTION_MD_BASE + 1)
#define OPTION_DIVIDE (OPTION_MD_BASE + 2)
+#define OPTION_MARCH (OPTION_MD_BASE + 3)
+#define OPTION_MTUNE (OPTION_MD_BASE + 4)
-struct option md_longopts[] = {
+struct option md_longopts[] =
+{
{"32", no_argument, NULL, OPTION_32},
-#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)
+#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) || defined(TE_PEP)
{"64", no_argument, NULL, OPTION_64},
#endif
{"divide", no_argument, NULL, OPTION_DIVIDE},
+ {"march", required_argument, NULL, OPTION_MARCH},
+ {"mtune", required_argument, NULL, OPTION_MTUNE},
{NULL, no_argument, NULL, 0}
};
size_t md_longopts_size = sizeof (md_longopts);
int
-md_parse_option (c, arg)
- int c;
- char *arg ATTRIBUTE_UNUSED;
+md_parse_option (int c, char *arg)
{
+ unsigned int i;
+
switch (c)
{
case 'n':
/* -s: On i386 Solaris, this tells the native assembler to use
.stab instead of .stab.excl. We always use .stab anyhow. */
break;
-
+#endif
+#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) || defined(TE_PEP)
case OPTION_64:
{
const char **list, **l;
list = bfd_target_list ();
for (l = list; *l != NULL; l++)
- if (strcmp (*l, "elf64-x86-64") == 0)
+ if (CONST_STRNEQ (*l, "elf64-x86-64")
+ || strcmp (*l, "coff-x86-64") == 0
+ || strcmp (*l, "pe-x86-64") == 0
+ || strcmp (*l, "pei-x86-64") == 0)
{
default_arch = "x86_64";
break;
#endif
break;
+ case OPTION_MARCH:
+ if (*arg == '.')
+ as_fatal (_("Invalid -march= option: `%s'"), arg);
+ for (i = 0; i < ARRAY_SIZE (cpu_arch); i++)
+ {
+ if (strcmp (arg, cpu_arch [i].name) == 0)
+ {
+ cpu_arch_isa = cpu_arch[i].type;
+ cpu_arch_isa_flags = cpu_arch[i].flags;
+ if (!cpu_arch_tune_set)
+ {
+ cpu_arch_tune = cpu_arch_isa;
+ cpu_arch_tune_flags = cpu_arch_isa_flags;
+ }
+ break;
+ }
+ }
+ if (i >= ARRAY_SIZE (cpu_arch))
+ as_fatal (_("Invalid -march= option: `%s'"), arg);
+ break;
+
+ case OPTION_MTUNE:
+ if (*arg == '.')
+ as_fatal (_("Invalid -mtune= option: `%s'"), arg);
+ for (i = 0; i < ARRAY_SIZE (cpu_arch); i++)
+ {
+ if (strcmp (arg, cpu_arch [i].name) == 0)
+ {
+ cpu_arch_tune_set = 1;
+ cpu_arch_tune = cpu_arch [i].type;
+ cpu_arch_tune_flags = cpu_arch[i].flags;
+ break;
+ }
+ }
+ if (i >= ARRAY_SIZE (cpu_arch))
+ as_fatal (_("Invalid -mtune= option: `%s'"), arg);
+ break;
+
default:
return 0;
}
fprintf (stream, _("\
-s ignored\n"));
#endif
+#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) || defined(TE_PEP)
+ fprintf (stream, _("\
+ --32/--64 generate 32bit/64bit code\n"));
+#endif
#ifdef SVR4_COMMENT_CHARS
fprintf (stream, _("\
--divide do not treat `/' as a comment character\n"));
fprintf (stream, _("\
--divide ignored\n"));
#endif
+ fprintf (stream, _("\
+ -march=CPU/-mtune=CPU generate code/optimize for CPU, where CPU is one of:\n\
+ i386, i486, pentium, pentiumpro, pentium4, nocona,\n\
+ core, core2, k6, athlon, k8, generic32, generic64\n"));
+
+}
+
+#if defined(TE_PEP)
+const char *
+x86_64_target_format (void)
+{
+ if (strcmp (default_arch, "x86_64") == 0)
+ {
+ set_code_flag (CODE_64BIT);
+ return COFF_TARGET_FORMAT;
+ }
+ else if (strcmp (default_arch, "i386") == 0)
+ {
+ set_code_flag (CODE_32BIT);
+ return "coff-i386";
+ }
+
+ as_fatal (_("Unknown architecture"));
+ return NULL;
}
+#endif
#if ((defined (OBJ_MAYBE_COFF) && defined (OBJ_MAYBE_AOUT)) \
|| defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF))
/* Pick the target format to use. */
const char *
-i386_target_format ()
+i386_target_format (void)
{
if (!strcmp (default_arch, "x86_64"))
- set_code_flag (CODE_64BIT);
+ {
+ set_code_flag (CODE_64BIT);
+ if (cpu_arch_isa_flags == 0)
+ cpu_arch_isa_flags = Cpu186|Cpu286|Cpu386|Cpu486
+ |Cpu586|Cpu686|CpuP4|CpuMMX|CpuMMX2
+ |CpuSSE|CpuSSE2;
+ if (cpu_arch_tune_flags == 0)
+ cpu_arch_tune_flags = Cpu186|Cpu286|Cpu386|Cpu486
+ |Cpu586|Cpu686|CpuP4|CpuMMX|CpuMMX2
+ |CpuSSE|CpuSSE2;
+ }
else if (!strcmp (default_arch, "i386"))
- set_code_flag (CODE_32BIT);
+ {
+ set_code_flag (CODE_32BIT);
+ if (cpu_arch_isa_flags == 0)
+ cpu_arch_isa_flags = Cpu186|Cpu286|Cpu386;
+ if (cpu_arch_tune_flags == 0)
+ cpu_arch_tune_flags = Cpu186|Cpu286|Cpu386;
+ }
else
as_fatal (_("Unknown architecture"));
switch (OUTPUT_FLAVOR)
object_64bit = 1;
use_rela_relocations = 1;
}
- return flag_code == CODE_64BIT ? "elf64-x86-64" : ELF_TARGET_FORMAT;
+ return flag_code == CODE_64BIT ? ELF_TARGET_FORMAT64 : ELF_TARGET_FORMAT;
}
#endif
default:
#endif /* OBJ_MAYBE_ more than one */
#if (defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF))
-void i386_elf_emit_arch_note ()
+void
+i386_elf_emit_arch_note (void)
{
if (IS_ELF && cpu_arch_name != NULL)
{
size, since the offset is always the last part of the insn. */
long
-md_pcrel_from (fixP)
- fixS *fixP;
+md_pcrel_from (fixS *fixP)
{
return fixP->fx_size + fixP->fx_where + fixP->fx_frag->fr_address;
}
#ifndef I386COFF
static void
-s_bss (ignore)
- int ignore ATTRIBUTE_UNUSED;
+s_bss (int ignore ATTRIBUTE_UNUSED)
{
int temp;
#endif
void
-i386_validate_fix (fixp)
- fixS *fixp;
+i386_validate_fix (fixS *fixp)
{
if (fixp->fx_subsy && fixp->fx_subsy == GOT_symbol)
{
#define T_SHR 15
/* Prototypes for intel parser functions. */
-static int intel_match_token PARAMS ((int code));
-static void intel_get_token PARAMS ((void));
-static void intel_putback_token PARAMS ((void));
-static int intel_expr PARAMS ((void));
-static int intel_e04 PARAMS ((void));
-static int intel_e05 PARAMS ((void));
-static int intel_e06 PARAMS ((void));
-static int intel_e09 PARAMS ((void));
-static int intel_bracket_expr PARAMS ((void));
-static int intel_e10 PARAMS ((void));
-static int intel_e11 PARAMS ((void));
+static int intel_match_token (int);
+static void intel_putback_token (void);
+static void intel_get_token (void);
+static int intel_expr (void);
+static int intel_e04 (void);
+static int intel_e05 (void);
+static int intel_e06 (void);
+static int intel_e09 (void);
+static int intel_e10 (void);
+static int intel_e11 (void);
static int
-i386_intel_operand (operand_string, got_a_float)
- char *operand_string;
- int got_a_float;
+i386_intel_operand (char *operand_string, int got_a_float)
{
int ret;
char *p;
expr' cmpOp e04 expr'
| Empty */
static int
-intel_expr ()
+intel_expr (void)
{
/* XXX Implement the comparison operators. */
return intel_e04 ();
e04' addOp e05 e04'
| Empty */
static int
-intel_e04 ()
+intel_e04 (void)
{
int nregs = -1;
e05' binOp e06 e05'
| Empty */
static int
-intel_e05 ()
+intel_e05 (void)
{
int nregs = ~NUM_ADDRESS_REGS;
e06' mulOp e09 e06'
| Empty */
static int
-intel_e06 ()
+intel_e06 (void)
{
int nregs = ~NUM_ADDRESS_REGS;
| : e10 e09'
| Empty */
static int
-intel_e09 ()
+intel_e09 (void)
{
int nregs = ~NUM_ADDRESS_REGS;
int in_offset = 0;
}
static int
-intel_bracket_expr ()
+intel_bracket_expr (void)
{
int was_offset = intel_parser.op_modifier & (1 << T_OFFSET);
const char *start = intel_parser.op_string;
e10' [ expr ] e10'
| Empty */
static int
-intel_e10 ()
+intel_e10 (void)
{
if (!intel_e11 ())
return 0;
| id
| constant */
static int
-intel_e11 ()
+intel_e11 (void)
{
switch (cur_token.code)
{
/* Match the given token against cur_token. If they match, read the next
token from the operand string. */
static int
-intel_match_token (code)
- int code;
+intel_match_token (int code)
{
if (cur_token.code == code)
{
/* Read a new token from intel_parser.op_string and store it in cur_token. */
static void
-intel_get_token ()
+intel_get_token (void)
{
char *end_op;
const reg_entry *reg;
/* Put cur_token back into the token stream and make cur_token point to
prev_token. */
static void
-intel_putback_token ()
+intel_putback_token (void)
{
if (cur_token.code != T_NIL)
{
}
int
-tc_x86_regname_to_dw2regnum (const char *regname)
+tc_x86_regname_to_dw2regnum (char *regname)
{
unsigned int regnum;
unsigned int regnames_count;
int
x86_64_section_word (char *str, size_t len)
{
- if (len == 5 && flag_code == CODE_64BIT && strncmp (str, "large", 5) == 0)
+ if (len == 5 && flag_code == CODE_64BIT && CONST_STRNEQ (str, "large"))
return SHF_X86_64_LARGE;
return -1;