/*
 * Derived from [gcc]/gcc/config/i386/i386.c
 * (pre-4.5 snapshot taken on 20091223)
 */


/* Subroutines used for code generation on IA-32.
   Copyright (C) 1988, 1992, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
   2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
   Free Software Foundation, Inc.

This file is part of GCC.

GCC is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3, or (at your option)
any later version.

GCC is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3.  If not see
<http://www.gnu.org/licenses/>.  */

#include "auto-host.h"
#ifndef ENABLE_BUILD_WITH_CXX
extern "C" {
#endif
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "tm.h"
#include "rtl.h"
#include "tree.h"
#include "tm_p.h"
#include "hard-reg-set.h"
#include "real.h"
#include "output.h"
#include "flags.h"
#include "except.h"
#include "function.h"
#include "toplev.h"
#include "basic-block.h"
#include "ggc.h"
#include "target.h"
#include "langhooks.h"
#include "cgraph.h"
#include "gimple.h"
#include "params.h"
#ifndef ENABLE_BUILD_WITH_CXX
} // extern "C"
#endif

#ifndef CHECK_STACK_LIMIT
#define CHECK_STACK_LIMIT (-1)
#endif

/* Return index of given mode in mult and division cost tables.  */
#define MODE_INDEX(mode)					\
  ((mode) == QImode ? 0						\
   : (mode) == HImode ? 1					\
   : (mode) == SImode ? 2					\
   : (mode) == DImode ? 3					\
   : 4)

/* Processor costs (relative to an add) */
/* We assume COSTS_N_INSNS is defined as (N)*4 and an addition is 2 bytes.  */
#define COSTS_N_BYTES(N) ((N) * 2)

#define DUMMY_STRINGOP_ALGS {libcall, {{-1, libcall}}}

const
struct processor_costs ix86_size_cost = {/* costs for tuning for size */
  COSTS_N_BYTES (2),			/* cost of an add instruction */
  COSTS_N_BYTES (3),			/* cost of a lea instruction */
  COSTS_N_BYTES (2),			/* variable shift costs */
  COSTS_N_BYTES (3),			/* constant shift costs */
  {COSTS_N_BYTES (3),			/* cost of starting multiply for QI */
   COSTS_N_BYTES (3),			/*                               HI */
   COSTS_N_BYTES (3),			/*                               SI */
   COSTS_N_BYTES (3),			/*                               DI */
   COSTS_N_BYTES (5)},			/*                            other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_BYTES (3),			/* cost of a divide/mod for QI */
   COSTS_N_BYTES (3),			/*                          HI */
   COSTS_N_BYTES (3),			/*                          SI */
   COSTS_N_BYTES (3),			/*                          DI */
   COSTS_N_BYTES (5)},			/*                       other */
  COSTS_N_BYTES (3),			/* cost of movsx */
  COSTS_N_BYTES (3),			/* cost of movzx */
  0,					/* "large" insn */
  2,					/* MOVE_RATIO */
  2,					/* cost for loading QImode using movzbl */
  {2, 2, 2},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 2, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {2, 2, 2},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {2, 2, 2},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  3,					/* cost of moving MMX register */
  {3, 3},				/* cost of loading MMX registers
					   in SImode and DImode */
  {3, 3},				/* cost of storing MMX registers
					   in SImode and DImode */
  3,					/* cost of moving SSE register */
  {3, 3, 3},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {3, 3, 3},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  0,					/* size of l1 cache  */
  0,					/* size of l2 cache  */
  0,					/* size of prefetch block */
  0,					/* number of parallel prefetches */
  2,					/* Branch cost */
  COSTS_N_BYTES (2),			/* cost of FADD and FSUB insns.  */
  COSTS_N_BYTES (2),			/* cost of FMUL instruction.  */
  COSTS_N_BYTES (2),			/* cost of FDIV instruction.  */
  COSTS_N_BYTES (2),			/* cost of FABS instruction.  */
  COSTS_N_BYTES (2),			/* cost of FCHS instruction.  */
  COSTS_N_BYTES (2),			/* cost of FSQRT instruction.  */
  {{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
   {rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}}},
  {{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
   {rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}}},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  1,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  1,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

/* Processor costs (relative to an add) */
static const
struct processor_costs i386_cost = {	/* 386 specific costs */
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (3),			/* variable shift costs */
  COSTS_N_INSNS (2),			/* constant shift costs */
  {COSTS_N_INSNS (6),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (6),			/*                               HI */
   COSTS_N_INSNS (6),			/*                               SI */
   COSTS_N_INSNS (6),			/*                               DI */
   COSTS_N_INSNS (6)},			/*                               other */
  COSTS_N_INSNS (1),			/* cost of multiply per each bit set */
  {COSTS_N_INSNS (23),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (23),			/*                          HI */
   COSTS_N_INSNS (23),			/*                          SI */
   COSTS_N_INSNS (23),			/*                          DI */
   COSTS_N_INSNS (23)},			/*                          other */
  COSTS_N_INSNS (3),			/* cost of movsx */
  COSTS_N_INSNS (2),			/* cost of movzx */
  15,					/* "large" insn */
  3,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {2, 4, 2},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 4, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {8, 8, 8},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {8, 8, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {4, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 8, 16},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 8, 16},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  0,					/* size of l1 cache  */
  0,					/* size of l2 cache  */
  0,					/* size of prefetch block */
  0,					/* number of parallel prefetches */
  1,					/* Branch cost */
  COSTS_N_INSNS (23),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (27),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (88),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (22),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (24),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (122),			/* cost of FSQRT instruction.  */
  {{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
   DUMMY_STRINGOP_ALGS},
  {{rep_prefix_1_byte, {{-1, rep_prefix_1_byte}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs i486_cost = {	/* 486 specific costs */
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (3),			/* variable shift costs */
  COSTS_N_INSNS (2),			/* constant shift costs */
  {COSTS_N_INSNS (12),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (12),			/*                               HI */
   COSTS_N_INSNS (12),			/*                               SI */
   COSTS_N_INSNS (12),			/*                               DI */
   COSTS_N_INSNS (12)},			/*                               other */
  1,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (40),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (40),			/*                          HI */
   COSTS_N_INSNS (40),			/*                          SI */
   COSTS_N_INSNS (40),			/*                          DI */
   COSTS_N_INSNS (40)},			/*                          other */
  COSTS_N_INSNS (3),			/* cost of movsx */
  COSTS_N_INSNS (2),			/* cost of movzx */
  15,					/* "large" insn */
  3,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {2, 4, 2},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 4, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {8, 8, 8},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {8, 8, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {4, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 8, 16},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 8, 16},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  4,					/* size of l1 cache.  486 has 8kB cache
					   shared for code and data, so 4kB is
					   not really precise.  */
  4,					/* size of l2 cache  */
  0,					/* size of prefetch block */
  0,					/* number of parallel prefetches */
  1,					/* Branch cost */
  COSTS_N_INSNS (8),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (16),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (73),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (3),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (3),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (83),			/* cost of FSQRT instruction.  */
  {{rep_prefix_4_byte, {{-1, rep_prefix_4_byte}}},
   DUMMY_STRINGOP_ALGS},
  {{rep_prefix_4_byte, {{-1, rep_prefix_4_byte}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs pentium_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (4),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (11),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (11),			/*                               HI */
   COSTS_N_INSNS (11),			/*                               SI */
   COSTS_N_INSNS (11),			/*                               DI */
   COSTS_N_INSNS (11)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (25),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (25),			/*                          HI */
   COSTS_N_INSNS (25),			/*                          SI */
   COSTS_N_INSNS (25),			/*                          DI */
   COSTS_N_INSNS (25)},			/*                          other */
  COSTS_N_INSNS (3),			/* cost of movsx */
  COSTS_N_INSNS (2),			/* cost of movzx */
  8,					/* "large" insn */
  6,					/* MOVE_RATIO */
  6,					/* cost for loading QImode using movzbl */
  {2, 4, 2},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 4, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {2, 2, 6},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 6},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  8,					/* cost of moving MMX register */
  {8, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {8, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 8, 16},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 8, 16},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  8,					/* size of l1 cache.  */
  8,					/* size of l2 cache  */
  0,					/* size of prefetch block */
  0,					/* number of parallel prefetches */
  2,					/* Branch cost */
  COSTS_N_INSNS (3),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (3),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (39),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (1),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (1),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (70),			/* cost of FSQRT instruction.  */
  {{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{-1, rep_prefix_4_byte}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs pentiumpro_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (4),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (4),			/*                               SI */
   COSTS_N_INSNS (4),			/*                               DI */
   COSTS_N_INSNS (4)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (17),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (17),			/*                          HI */
   COSTS_N_INSNS (17),			/*                          SI */
   COSTS_N_INSNS (17),			/*                          DI */
   COSTS_N_INSNS (17)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  6,					/* MOVE_RATIO */
  2,					/* cost for loading QImode using movzbl */
  {4, 4, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 2, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {2, 2, 6},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 6},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {2, 2},				/* cost of loading MMX registers
					   in SImode and DImode */
  {2, 2},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {2, 2, 8},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {2, 2, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  8,					/* size of l1 cache.  */
  256,					/* size of l2 cache  */
  32,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  2,					/* Branch cost */
  COSTS_N_INSNS (3),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (5),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (56),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (56),			/* cost of FSQRT instruction.  */
  /* PentiumPro has optimized rep instructions for blocks aligned by 8 bytes (we ensure
     the alignment).  For small blocks inline loop is still a noticeable win, for bigger
     blocks either rep movsl or rep movsb is way to go.  Rep movsb has apparently
     more expensive startup time in CPU, but after 4K the difference is down in the noise.
   */
  {{rep_prefix_4_byte, {{128, loop}, {1024, unrolled_loop},
			{8192, rep_prefix_4_byte}, {-1, rep_prefix_1_byte}}},
   DUMMY_STRINGOP_ALGS},
  {{rep_prefix_4_byte, {{1024, unrolled_loop},
  		        {8192, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs geode_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (2),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (7),			/*                               SI */
   COSTS_N_INSNS (7),			/*                               DI */
   COSTS_N_INSNS (7)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (15),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (23),			/*                          HI */
   COSTS_N_INSNS (39),			/*                          SI */
   COSTS_N_INSNS (39),			/*                          DI */
   COSTS_N_INSNS (39)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  4,					/* MOVE_RATIO */
  1,					/* cost for loading QImode using movzbl */
  {1, 1, 1},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {1, 1, 1},				/* cost of storing integer registers */
  1,					/* cost of reg,reg fld/fst */
  {1, 1, 1},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 6, 6},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */

  1,					/* cost of moving MMX register */
  {1, 1},				/* cost of loading MMX registers
					   in SImode and DImode */
  {1, 1},				/* cost of storing MMX registers
					   in SImode and DImode */
  1,					/* cost of moving SSE register */
  {1, 1, 1},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {1, 1, 1},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  1,					/* MMX or SSE register to integer */
  64,					/* size of l1 cache.  */
  128,					/* size of l2 cache.  */
  32,					/* size of prefetch block */
  1,					/* number of parallel prefetches */
  1,					/* Branch cost */
  COSTS_N_INSNS (6),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (11),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (47),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (1),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (1),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (54),			/* cost of FSQRT instruction.  */
  {{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs k6_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (2),			/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (3),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (3),			/*                               DI */
   COSTS_N_INSNS (3)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (18),			/*                          HI */
   COSTS_N_INSNS (18),			/*                          SI */
   COSTS_N_INSNS (18),			/*                          DI */
   COSTS_N_INSNS (18)},			/*                          other */
  COSTS_N_INSNS (2),			/* cost of movsx */
  COSTS_N_INSNS (2),			/* cost of movzx */
  8,					/* "large" insn */
  4,					/* MOVE_RATIO */
  3,					/* cost for loading QImode using movzbl */
  {4, 5, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 3, 2},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {6, 6, 6},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 4},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {2, 2},				/* cost of loading MMX registers
					   in SImode and DImode */
  {2, 2},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {2, 2, 8},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {2, 2, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  6,					/* MMX or SSE register to integer */
  32,					/* size of l1 cache.  */
  32,					/* size of l2 cache.  Some models
					   have integrated l2 cache, but
					   optimizing for k6 is not important
					   enough to worry about that.  */
  32,					/* size of prefetch block */
  1,					/* number of parallel prefetches */
  1,					/* Branch cost */
  COSTS_N_INSNS (2),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (2),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (56),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (56),			/* cost of FSQRT instruction.  */
  {{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{256, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs athlon_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (2),			/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (5),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (5),			/*                               HI */
   COSTS_N_INSNS (5),			/*                               SI */
   COSTS_N_INSNS (5),			/*                               DI */
   COSTS_N_INSNS (5)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (26),			/*                          HI */
   COSTS_N_INSNS (42),			/*                          SI */
   COSTS_N_INSNS (74),			/*                          DI */
   COSTS_N_INSNS (74)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  9,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {3, 4, 3},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {3, 4, 3},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {4, 4, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {4, 4},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 4},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 4, 6},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 4, 5},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  5,					/* MMX or SSE register to integer */
  64,					/* size of l1 cache.  */
  256,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  5,					/* Branch cost */
  COSTS_N_INSNS (4),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (4),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (24),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (35),			/* cost of FSQRT instruction.  */
  /* For some reason, Athlon deals better with REP prefix (relative to loops)
     compared to K8. Alignment becomes important after 8 bytes for memcpy and
     128 bytes for memset.  */
  {{libcall, {{2048, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{2048, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs k8_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (2),			/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (4),			/*                               DI */
   COSTS_N_INSNS (5)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (26),			/*                          HI */
   COSTS_N_INSNS (42),			/*                          SI */
   COSTS_N_INSNS (74),			/*                          DI */
   COSTS_N_INSNS (74)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  9,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {3, 4, 3},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {3, 4, 3},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {4, 4, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {3, 3},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 4},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 3, 6},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 4, 5},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  5,					/* MMX or SSE register to integer */
  64,					/* size of l1 cache.  */
  512,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  /* New AMD processors never drop prefetches; if they cannot be performed
     immediately, they are queued.  We set number of simultaneous prefetches
     to a large constant to reflect this (it probably is not a good idea not
     to limit number of prefetches at all, as their execution also takes some
     time).  */
  100,					/* number of parallel prefetches */
  3,					/* Branch cost */
  COSTS_N_INSNS (4),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (4),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (19),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (35),			/* cost of FSQRT instruction.  */
  /* K8 has optimized REP instruction for medium sized blocks, but for very small
     blocks it is better to use loop. For large blocks, libcall can do
     nontemporary accesses and beat inline considerably.  */
  {{libcall, {{6, loop}, {14, unrolled_loop}, {-1, rep_prefix_4_byte}}},
   {libcall, {{16, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  {{libcall, {{8, loop}, {24, unrolled_loop},
	      {2048, rep_prefix_4_byte}, {-1, libcall}}},
   {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  4,                                    /* scalar_stmt_cost.  */
  2,                                    /* scalar load_cost.  */
  2,                                    /* scalar_store_cost.  */
  5,                                    /* vec_stmt_cost.  */
  0,                                    /* vec_to_scalar_cost.  */
  2,                                    /* scalar_to_vec_cost.  */
  2,                                    /* vec_align_load_cost.  */
  3,                                    /* vec_unalign_load_cost.  */
  3,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  2,                                    /* cond_not_taken_branch_cost.  */
};

struct processor_costs amdfam10_cost = {
  COSTS_N_INSNS (1),                    /* cost of an add instruction */
  COSTS_N_INSNS (2),                    /* cost of a lea instruction */
  COSTS_N_INSNS (1),                    /* variable shift costs */
  COSTS_N_INSNS (1),                    /* constant shift costs */
  {COSTS_N_INSNS (3),                   /* cost of starting multiply for QI */
   COSTS_N_INSNS (4),                   /*                               HI */
   COSTS_N_INSNS (3),                   /*                               SI */
   COSTS_N_INSNS (4),                   /*                               DI */
   COSTS_N_INSNS (5)},                  /*                               other */
  0,                                    /* cost of multiply per each bit set */
  {COSTS_N_INSNS (19),                  /* cost of a divide/mod for QI */
   COSTS_N_INSNS (35),                  /*                          HI */
   COSTS_N_INSNS (51),                  /*                          SI */
   COSTS_N_INSNS (83),                  /*                          DI */
   COSTS_N_INSNS (83)},                 /*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  9,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {3, 4, 3},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {3, 4, 3},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {4, 4, 12},				/* cost of loading fp registers
		   			   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
 		   			   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {3, 3},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 4},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {4, 4, 3},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 4, 5},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  3,					/* MMX or SSE register to integer */
  					/* On K8
  					    MOVD reg64, xmmreg 	Double	FSTORE 4
					    MOVD reg32, xmmreg 	Double	FSTORE 4
					   On AMDFAM10
					    MOVD reg64, xmmreg 	Double	FADD 3
                                                                1/1  1/1
					    MOVD reg32, xmmreg 	Double	FADD 3
                                                                1/1  1/1 */
  64,					/* size of l1 cache.  */
  512,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  /* New AMD processors never drop prefetches; if they cannot be performed
     immediately, they are queued.  We set number of simultaneous prefetches
     to a large constant to reflect this (it probably is not a good idea not
     to limit number of prefetches at all, as their execution also takes some
     time).  */
  100,					/* number of parallel prefetches */
  2,					/* Branch cost */
  COSTS_N_INSNS (4),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (4),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (19),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (35),			/* cost of FSQRT instruction.  */

  /* AMDFAM10 has optimized REP instruction for medium sized blocks, but for
     very small blocks it is better to use loop. For large blocks, libcall can
     do nontemporary accesses and beat inline considerably.  */
  {{libcall, {{6, loop}, {14, unrolled_loop}, {-1, rep_prefix_4_byte}}},
   {libcall, {{16, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  {{libcall, {{8, loop}, {24, unrolled_loop},
	      {2048, rep_prefix_4_byte}, {-1, libcall}}},
   {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  4,                                    /* scalar_stmt_cost.  */
  2,                                    /* scalar load_cost.  */
  2,                                    /* scalar_store_cost.  */
  6,                                    /* vec_stmt_cost.  */
  0,                                    /* vec_to_scalar_cost.  */
  2,                                    /* scalar_to_vec_cost.  */
  2,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  2,                                    /* vec_store_cost.  */
  2,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs pentium4_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (3),			/* cost of a lea instruction */
  COSTS_N_INSNS (4),			/* variable shift costs */
  COSTS_N_INSNS (4),			/* constant shift costs */
  {COSTS_N_INSNS (15),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (15),			/*                               HI */
   COSTS_N_INSNS (15),			/*                               SI */
   COSTS_N_INSNS (15),			/*                               DI */
   COSTS_N_INSNS (15)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (56),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (56),			/*                          HI */
   COSTS_N_INSNS (56),			/*                          SI */
   COSTS_N_INSNS (56),			/*                          DI */
   COSTS_N_INSNS (56)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  16,					/* "large" insn */
  6,					/* MOVE_RATIO */
  2,					/* cost for loading QImode using movzbl */
  {4, 5, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {2, 3, 2},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {2, 2, 6},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 6},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {2, 2},				/* cost of loading MMX registers
					   in SImode and DImode */
  {2, 2},				/* cost of storing MMX registers
					   in SImode and DImode */
  12,					/* cost of moving SSE register */
  {12, 12, 12},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {2, 2, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  10,					/* MMX or SSE register to integer */
  8,					/* size of l1 cache.  */
  256,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  2,					/* Branch cost */
  COSTS_N_INSNS (5),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (7),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (43),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (43),			/* cost of FSQRT instruction.  */
  {{libcall, {{12, loop_1_byte}, {-1, rep_prefix_4_byte}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{6, loop_1_byte}, {48, loop}, {20480, rep_prefix_4_byte},
   {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs nocona_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1),			/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (10),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (10),			/*                               HI */
   COSTS_N_INSNS (10),			/*                               SI */
   COSTS_N_INSNS (10),			/*                               DI */
   COSTS_N_INSNS (10)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (66),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (66),			/*                          HI */
   COSTS_N_INSNS (66),			/*                          SI */
   COSTS_N_INSNS (66),			/*                          DI */
   COSTS_N_INSNS (66)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  16,					/* "large" insn */
  17,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {4, 4, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {4, 4, 4},				/* cost of storing integer registers */
  3,					/* cost of reg,reg fld/fst */
  {12, 12, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 4},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  6,					/* cost of moving MMX register */
  {12, 12},				/* cost of loading MMX registers
					   in SImode and DImode */
  {12, 12},				/* cost of storing MMX registers
					   in SImode and DImode */
  6,					/* cost of moving SSE register */
  {12, 12, 12},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {12, 12, 12},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  8,					/* MMX or SSE register to integer */
  8,					/* size of l1 cache.  */
  1024,					/* size of l2 cache.  */
  128,					/* size of prefetch block */
  8,					/* number of parallel prefetches */
  1,					/* Branch cost */
  COSTS_N_INSNS (6),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (8),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (40),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (3),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (3),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (44),			/* cost of FSQRT instruction.  */
  {{libcall, {{12, loop_1_byte}, {-1, rep_prefix_4_byte}}},
   {libcall, {{32, loop}, {20000, rep_prefix_8_byte},
	      {100000, unrolled_loop}, {-1, libcall}}}},
  {{libcall, {{6, loop_1_byte}, {48, loop}, {20480, rep_prefix_4_byte},
   {-1, libcall}}},
   {libcall, {{24, loop}, {64, unrolled_loop},
	      {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs core2_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1) + 1,		/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (3),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (3),			/*                               DI */
   COSTS_N_INSNS (3)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (22),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (22),			/*                          HI */
   COSTS_N_INSNS (22),			/*                          SI */
   COSTS_N_INSNS (22),			/*                          DI */
   COSTS_N_INSNS (22)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  16,					/* MOVE_RATIO */
  2,					/* cost for loading QImode using movzbl */
  {6, 6, 6},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {4, 4, 4},				/* cost of storing integer registers */
  2,					/* cost of reg,reg fld/fst */
  {6, 6, 6},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {4, 4, 4},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {6, 6},				/* cost of loading MMX registers
					   in SImode and DImode */
  {4, 4},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {6, 6, 6},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {4, 4, 4},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  2,					/* MMX or SSE register to integer */
  32,					/* size of l1 cache.  */
  2048,					/* size of l2 cache.  */
  128,					/* size of prefetch block */
  8,					/* number of parallel prefetches */
  3,					/* Branch cost */
  COSTS_N_INSNS (3),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (5),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (32),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (1),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (1),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (58),			/* cost of FSQRT instruction.  */
  {{libcall, {{11, loop}, {-1, rep_prefix_4_byte}}},
   {libcall, {{32, loop}, {64, rep_prefix_4_byte},
	      {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  {{libcall, {{8, loop}, {15, unrolled_loop},
	      {2048, rep_prefix_4_byte}, {-1, libcall}}},
   {libcall, {{24, loop}, {32, unrolled_loop},
	      {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

static const
struct processor_costs atom_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1) + 1,		/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (4),			/*                               DI */
   COSTS_N_INSNS (2)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (26),			/*                          HI */
   COSTS_N_INSNS (42),			/*                          SI */
   COSTS_N_INSNS (74),			/*                          DI */
   COSTS_N_INSNS (74)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  17,					/* MOVE_RATIO */
  2,					/* cost for loading QImode using movzbl */
  {4, 4, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {4, 4, 4},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {12, 12, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {8, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {8, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {8, 8, 8},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {8, 8, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  5,					/* MMX or SSE register to integer */
  32,					/* size of l1 cache.  */
  256,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  3,					/* Branch cost */
  COSTS_N_INSNS (8),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (8),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (20),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (8),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (8),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (40),			/* cost of FSQRT instruction.  */
  {{libcall, {{11, loop}, {-1, rep_prefix_4_byte}}},
   {libcall, {{32, loop}, {64, rep_prefix_4_byte},
          {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  {{libcall, {{8, loop}, {15, unrolled_loop},
          {2048, rep_prefix_4_byte}, {-1, libcall}}},
   {libcall, {{24, loop}, {32, unrolled_loop},
          {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

/* Generic64 should produce code tuned for Nocona and K8.  */
static const
struct processor_costs generic64_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  /* On all chips taken into consideration lea is 2 cycles and more.  With
     this cost however our current implementation of synth_mult results in
     use of unnecessary temporary registers causing regression on several
     SPECfp benchmarks.  */
  COSTS_N_INSNS (1) + 1,		/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (4),			/*                               DI */
   COSTS_N_INSNS (2)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (26),			/*                          HI */
   COSTS_N_INSNS (42),			/*                          SI */
   COSTS_N_INSNS (74),			/*                          DI */
   COSTS_N_INSNS (74)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  17,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {4, 4, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {4, 4, 4},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {12, 12, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {8, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {8, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {8, 8, 8},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {8, 8, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  5,					/* MMX or SSE register to integer */
  32,					/* size of l1 cache.  */
  512,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  /* Benchmarks shows large regressions on K8 sixtrack benchmark when this value
     is increased to perhaps more appropriate value of 5.  */
  3,					/* Branch cost */
  COSTS_N_INSNS (8),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (8),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (20),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (8),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (8),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (40),			/* cost of FSQRT instruction.  */
  {DUMMY_STRINGOP_ALGS,
   {libcall, {{32, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  {DUMMY_STRINGOP_ALGS,
   {libcall, {{32, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

/* Generic32 should produce code tuned for Athlon, PPro, Pentium4, Nocona and K8.  */
static const
struct processor_costs generic32_cost = {
  COSTS_N_INSNS (1),			/* cost of an add instruction */
  COSTS_N_INSNS (1) + 1,		/* cost of a lea instruction */
  COSTS_N_INSNS (1),			/* variable shift costs */
  COSTS_N_INSNS (1),			/* constant shift costs */
  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
   COSTS_N_INSNS (4),			/*                               HI */
   COSTS_N_INSNS (3),			/*                               SI */
   COSTS_N_INSNS (4),			/*                               DI */
   COSTS_N_INSNS (2)},			/*                               other */
  0,					/* cost of multiply per each bit set */
  {COSTS_N_INSNS (18),			/* cost of a divide/mod for QI */
   COSTS_N_INSNS (26),			/*                          HI */
   COSTS_N_INSNS (42),			/*                          SI */
   COSTS_N_INSNS (74),			/*                          DI */
   COSTS_N_INSNS (74)},			/*                          other */
  COSTS_N_INSNS (1),			/* cost of movsx */
  COSTS_N_INSNS (1),			/* cost of movzx */
  8,					/* "large" insn */
  17,					/* MOVE_RATIO */
  4,					/* cost for loading QImode using movzbl */
  {4, 4, 4},				/* cost of loading integer registers
					   in QImode, HImode and SImode.
					   Relative to reg-reg move (2).  */
  {4, 4, 4},				/* cost of storing integer registers */
  4,					/* cost of reg,reg fld/fst */
  {12, 12, 12},				/* cost of loading fp registers
					   in SFmode, DFmode and XFmode */
  {6, 6, 8},				/* cost of storing fp registers
					   in SFmode, DFmode and XFmode */
  2,					/* cost of moving MMX register */
  {8, 8},				/* cost of loading MMX registers
					   in SImode and DImode */
  {8, 8},				/* cost of storing MMX registers
					   in SImode and DImode */
  2,					/* cost of moving SSE register */
  {8, 8, 8},				/* cost of loading SSE registers
					   in SImode, DImode and TImode */
  {8, 8, 8},				/* cost of storing SSE registers
					   in SImode, DImode and TImode */
  5,					/* MMX or SSE register to integer */
  32,					/* size of l1 cache.  */
  256,					/* size of l2 cache.  */
  64,					/* size of prefetch block */
  6,					/* number of parallel prefetches */
  3,					/* Branch cost */
  COSTS_N_INSNS (8),			/* cost of FADD and FSUB insns.  */
  COSTS_N_INSNS (8),			/* cost of FMUL instruction.  */
  COSTS_N_INSNS (20),			/* cost of FDIV instruction.  */
  COSTS_N_INSNS (8),			/* cost of FABS instruction.  */
  COSTS_N_INSNS (8),			/* cost of FCHS instruction.  */
  COSTS_N_INSNS (40),			/* cost of FSQRT instruction.  */
  {{libcall, {{32, loop}, {8192, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  {{libcall, {{32, loop}, {8192, rep_prefix_4_byte}, {-1, libcall}}},
   DUMMY_STRINGOP_ALGS},
  1,                                    /* scalar_stmt_cost.  */
  1,                                    /* scalar load_cost.  */
  1,                                    /* scalar_store_cost.  */
  1,                                    /* vec_stmt_cost.  */
  1,                                    /* vec_to_scalar_cost.  */
  1,                                    /* scalar_to_vec_cost.  */
  1,                                    /* vec_align_load_cost.  */
  2,                                    /* vec_unalign_load_cost.  */
  1,                                    /* vec_store_cost.  */
  3,                                    /* cond_taken_branch_cost.  */
  1,                                    /* cond_not_taken_branch_cost.  */
};

const struct processor_costs *ix86_cost = &pentium_cost;

/* Processor feature/optimization bitmasks.  */
#define m_386 (1<<PROCESSOR_I386)
#define m_486 (1<<PROCESSOR_I486)
#define m_PENT (1<<PROCESSOR_PENTIUM)
#define m_PPRO (1<<PROCESSOR_PENTIUMPRO)
#define m_PENT4  (1<<PROCESSOR_PENTIUM4)
#define m_NOCONA  (1<<PROCESSOR_NOCONA)
#define m_CORE2  (1<<PROCESSOR_CORE2)
#define m_ATOM  (1<<PROCESSOR_ATOM)

#define m_GEODE  (1<<PROCESSOR_GEODE)
#define m_K6  (1<<PROCESSOR_K6)
#define m_K6_GEODE  (m_K6 | m_GEODE)
#define m_K8  (1<<PROCESSOR_K8)
#define m_ATHLON  (1<<PROCESSOR_ATHLON)
#define m_ATHLON_K8  (m_K8 | m_ATHLON)
#define m_AMDFAM10  (1<<PROCESSOR_AMDFAM10)
#define m_AMD_MULTIPLE  (m_K8 | m_ATHLON | m_AMDFAM10)

#define m_GENERIC32 (1<<PROCESSOR_GENERIC32)
#define m_GENERIC64 (1<<PROCESSOR_GENERIC64)

/* Generic instruction choice should be common subset of supported CPUs
   (PPro/PENT4/NOCONA/CORE2/Athlon/K8).  */
#define m_GENERIC (m_GENERIC32 | m_GENERIC64)

/* In case the average insn count for single function invocation is
   lower than this constant, emit fast (but longer) prologue and
   epilogue code.  */
#define FAST_PROLOGUE_INSN_COUNT 20

/* Names for 8 (low), 8 (high), and 16-bit registers, respectively.  */
static const char *const qi_reg_name[] = QI_REGISTER_NAMES;
static const char *const qi_high_reg_name[] = QI_HIGH_REGISTER_NAMES;
static const char *const hi_reg_name[] = HI_REGISTER_NAMES;

/* Array of the smallest class containing reg number REGNO, indexed by
   REGNO.  Used by REGNO_REG_CLASS in i386.h.  */

enum reg_class const regclass_map[FIRST_PSEUDO_REGISTER] =
{
  /* ax, dx, cx, bx */
  AREG, DREG, CREG, BREG,
  /* si, di, bp, sp */
  SIREG, DIREG, NON_Q_REGS, NON_Q_REGS,
  /* FP registers */
  FP_TOP_REG, FP_SECOND_REG, FLOAT_REGS, FLOAT_REGS,
  FLOAT_REGS, FLOAT_REGS, FLOAT_REGS, FLOAT_REGS,
  /* arg pointer */
  NON_Q_REGS,
  /* flags, fpsr, fpcr, frame */
  NO_REGS, NO_REGS, NO_REGS, NON_Q_REGS,
  /* SSE registers */
  SSE_FIRST_REG, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS,
  SSE_REGS, SSE_REGS,
  /* MMX registers */
  MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS, MMX_REGS,
  MMX_REGS, MMX_REGS,
  /* REX registers */
  NON_Q_REGS, NON_Q_REGS, NON_Q_REGS, NON_Q_REGS,
  NON_Q_REGS, NON_Q_REGS, NON_Q_REGS, NON_Q_REGS,
  /* SSE REX registers */
  SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS, SSE_REGS,
  SSE_REGS, SSE_REGS,
};

/* The "default" register map used in 32bit mode.  */

int const dbx_register_map[FIRST_PSEUDO_REGISTER] =
{
  0, 2, 1, 3, 6, 7, 4, 5,		/* general regs */
  12, 13, 14, 15, 16, 17, 18, 19,	/* fp regs */
  -1, -1, -1, -1, -1,			/* arg, flags, fpsr, fpcr, frame */
  21, 22, 23, 24, 25, 26, 27, 28,	/* SSE */
  29, 30, 31, 32, 33, 34, 35, 36,       /* MMX */
  -1, -1, -1, -1, -1, -1, -1, -1,	/* extended integer registers */
  -1, -1, -1, -1, -1, -1, -1, -1,	/* extended SSE registers */
};

/* The "default" register map used in 64bit mode.  */

int const dbx64_register_map[FIRST_PSEUDO_REGISTER] =
{
  0, 1, 2, 3, 4, 5, 6, 7,		/* general regs */
  33, 34, 35, 36, 37, 38, 39, 40,	/* fp regs */
  -1, -1, -1, -1, -1,			/* arg, flags, fpsr, fpcr, frame */
  17, 18, 19, 20, 21, 22, 23, 24,	/* SSE */
  41, 42, 43, 44, 45, 46, 47, 48,       /* MMX */
  8,9,10,11,12,13,14,15,		/* extended integer registers */
  25, 26, 27, 28, 29, 30, 31, 32,	/* extended SSE registers */
};

/* Define the register numbers to be used in Dwarf debugging information.
   The SVR4 reference port C compiler uses the following register numbers
   in its Dwarf output code:
	0 for %eax (gcc regno = 0)
	1 for %ecx (gcc regno = 2)
	2 for %edx (gcc regno = 1)
	3 for %ebx (gcc regno = 3)
	4 for %esp (gcc regno = 7)
	5 for %ebp (gcc regno = 6)
	6 for %esi (gcc regno = 4)
	7 for %edi (gcc regno = 5)
   The following three DWARF register numbers are never generated by
   the SVR4 C compiler or by the GNU compilers, but SDB on x86/svr4
   believes these numbers have these meanings.
	8  for %eip    (no gcc equivalent)
	9  for %eflags (gcc regno = 17)
	10 for %trapno (no gcc equivalent)
   It is not at all clear how we should number the FP stack registers
   for the x86 architecture.  If the version of SDB on x86/svr4 were
   a bit less brain dead with respect to floating-point then we would
   have a precedent to follow with respect to DWARF register numbers
   for x86 FP registers, but the SDB on x86/svr4 is so completely
   broken with respect to FP registers that it is hardly worth thinking
   of it as something to strive for compatibility with.
   The version of x86/svr4 SDB I have at the moment does (partially)
   seem to believe that DWARF register number 11 is associated with
   the x86 register %st(0), but that's about all.  Higher DWARF
   register numbers don't seem to be associated with anything in
   particular, and even for DWARF regno 11, SDB only seems to under-
   stand that it should say that a variable lives in %st(0) (when
   asked via an `=' command) if we said it was in DWARF regno 11,
   but SDB still prints garbage when asked for the value of the
   variable in question (via a `/' command).
   (Also note that the labels SDB prints for various FP stack regs
   when doing an `x' command are all wrong.)
   Note that these problems generally don't affect the native SVR4
   C compiler because it doesn't allow the use of -O with -g and
   because when it is *not* optimizing, it allocates a memory
   location for each floating-point variable, and the memory
   location is what gets described in the DWARF AT_location
   attribute for the variable in question.
   Regardless of the severe mental illness of the x86/svr4 SDB, we
   do something sensible here and we use the following DWARF
   register numbers.  Note that these are all stack-top-relative
   numbers.
	11 for %st(0) (gcc regno = 8)
	12 for %st(1) (gcc regno = 9)
	13 for %st(2) (gcc regno = 10)
	14 for %st(3) (gcc regno = 11)
	15 for %st(4) (gcc regno = 12)
	16 for %st(5) (gcc regno = 13)
	17 for %st(6) (gcc regno = 14)
	18 for %st(7) (gcc regno = 15)
*/
int const svr4_dbx_register_map[FIRST_PSEUDO_REGISTER] =
{
  0, 2, 1, 3, 6, 7, 5, 4,		/* general regs */
  11, 12, 13, 14, 15, 16, 17, 18,	/* fp regs */
  -1, 9, -1, -1, -1,			/* arg, flags, fpsr, fpcr, frame */
  21, 22, 23, 24, 25, 26, 27, 28,	/* SSE registers */
  29, 30, 31, 32, 33, 34, 35, 36,	/* MMX registers */
  -1, -1, -1, -1, -1, -1, -1, -1,	/* extended integer registers */
  -1, -1, -1, -1, -1, -1, -1, -1,	/* extended SSE registers */
};

/* Test and compare insns in i386.md store the information needed to
   generate branch and scc insns here.  */

rtx ix86_compare_op0 = NULL_RTX;
rtx ix86_compare_op1 = NULL_RTX;

/* Define parameter passing and return registers.  */

static int const x86_64_int_parameter_registers[6] =
{
  DI_REG, SI_REG, DX_REG, CX_REG, R8_REG, R9_REG
};

static int const x86_64_ms_abi_int_parameter_registers[4] =
{
  CX_REG, DX_REG, R8_REG, R9_REG
};

static int const x86_64_int_return_registers[4] =
{
  AX_REG, DX_REG, DI_REG, SI_REG
};

/* Define the structure for the machine field in struct function.  */

struct GTY(()) stack_local_entry {
  unsigned short mode;
  unsigned short n;
  rtx rtl;
  struct stack_local_entry *next;
};

/* Structure describing stack frame layout.
   Stack grows downward:

   [arguments]
					      <- ARG_POINTER
   saved pc

   saved frame pointer if frame_pointer_needed
					      <- HARD_FRAME_POINTER
   [saved regs]

   [padding0]

   [saved SSE regs]

   [padding1]          \
		        )
   [va_arg registers]  (
		        > to_allocate	      <- FRAME_POINTER
   [frame]	       (
		        )
   [padding2]	       /
  */
struct ix86_frame
{
  int padding0;
  int nsseregs;
  int nregs;
  int padding1;
  int va_arg_size;
  HOST_WIDE_INT frame;
  int padding2;
  int outgoing_arguments_size;
  int red_zone_size;

  HOST_WIDE_INT to_allocate;
  /* The offsets relative to ARG_POINTER.  */
  HOST_WIDE_INT frame_pointer_offset;
  HOST_WIDE_INT hard_frame_pointer_offset;
  HOST_WIDE_INT stack_pointer_offset;

  /* When save_regs_using_mov is set, emit prologue using
     move instead of push instructions.  */
  bool save_regs_using_mov;
};

/* Which cpu are we optimizing for.  */
enum processor_type ix86_tune;

/* Which instruction set architecture to use.  */
enum processor_type ix86_arch;

/* Preferred alignment for stack boundary in bits.  */
unsigned int ix86_preferred_stack_boundary;

/* Prefix built by ASM_GENERATE_INTERNAL_LABEL.  */
char internal_label_prefix[16];
int internal_label_prefix_len;

/* Fence to use after loop using movnt.  */
tree x86_mfence;

#define MAX_CLASSES 4


enum ix86_function_specific_strings
{
  IX86_FUNCTION_SPECIFIC_ARCH,
  IX86_FUNCTION_SPECIFIC_TUNE,
  IX86_FUNCTION_SPECIFIC_FPMATH,
  IX86_FUNCTION_SPECIFIC_MAX
};


/* The svr4 ABI for the i386 says that records and unions are returned
   in memory.  */
#ifndef DEFAULT_PCC_STRUCT_RETURN
#define DEFAULT_PCC_STRUCT_RETURN 1
#endif

/* Define a set of ISAs which are available when a given ISA is
   enabled.  MMX and SSE ISAs are handled separately.  */

#define OPTION_MASK_ISA_MMX_SET OPTION_MASK_ISA_MMX
#define OPTION_MASK_ISA_3DNOW_SET \
  (OPTION_MASK_ISA_3DNOW | OPTION_MASK_ISA_MMX_SET)

#define OPTION_MASK_ISA_SSE_SET OPTION_MASK_ISA_SSE
#define OPTION_MASK_ISA_SSE2_SET \
  (OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE_SET)
#define OPTION_MASK_ISA_SSE3_SET \
  (OPTION_MASK_ISA_SSE3 | OPTION_MASK_ISA_SSE2_SET)
#define OPTION_MASK_ISA_SSSE3_SET \
  (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSE3_SET)
#define OPTION_MASK_ISA_SSE4_1_SET \
  (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSSE3_SET)
#define OPTION_MASK_ISA_SSE4_2_SET \
  (OPTION_MASK_ISA_SSE4_2 | OPTION_MASK_ISA_SSE4_1_SET)
#define OPTION_MASK_ISA_AVX_SET \
  (OPTION_MASK_ISA_AVX | OPTION_MASK_ISA_SSE4_2_SET)
#define OPTION_MASK_ISA_FMA_SET \
  (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_AVX_SET)

/* SSE4 includes both SSE4.1 and SSE4.2. -msse4 should be the same
   as -msse4.2.  */
#define OPTION_MASK_ISA_SSE4_SET OPTION_MASK_ISA_SSE4_2_SET

#define OPTION_MASK_ISA_SSE4A_SET \
  (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_SSE3_SET)
#define OPTION_MASK_ISA_FMA4_SET \
  (OPTION_MASK_ISA_FMA4 | OPTION_MASK_ISA_SSE4A_SET \
   | OPTION_MASK_ISA_AVX_SET)
#define OPTION_MASK_ISA_XOP_SET \
  (OPTION_MASK_ISA_XOP | OPTION_MASK_ISA_FMA4_SET)
#define OPTION_MASK_ISA_LWP_SET \
  OPTION_MASK_ISA_LWP

/* AES and PCLMUL need SSE2 because they use xmm registers */
#define OPTION_MASK_ISA_AES_SET \
  (OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2_SET)
#define OPTION_MASK_ISA_PCLMUL_SET \
  (OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2_SET)

#define OPTION_MASK_ISA_ABM_SET \
  (OPTION_MASK_ISA_ABM | OPTION_MASK_ISA_POPCNT)

#define OPTION_MASK_ISA_POPCNT_SET OPTION_MASK_ISA_POPCNT
#define OPTION_MASK_ISA_CX16_SET OPTION_MASK_ISA_CX16
#define OPTION_MASK_ISA_SAHF_SET OPTION_MASK_ISA_SAHF
#define OPTION_MASK_ISA_MOVBE_SET OPTION_MASK_ISA_MOVBE
#define OPTION_MASK_ISA_CRC32_SET OPTION_MASK_ISA_CRC32

/* Define a set of ISAs which aren't available when a given ISA is
   disabled.  MMX and SSE ISAs are handled separately.  */

#define OPTION_MASK_ISA_MMX_UNSET \
  (OPTION_MASK_ISA_MMX | OPTION_MASK_ISA_3DNOW_UNSET)
#define OPTION_MASK_ISA_3DNOW_UNSET \
  (OPTION_MASK_ISA_3DNOW | OPTION_MASK_ISA_3DNOW_A_UNSET)
#define OPTION_MASK_ISA_3DNOW_A_UNSET OPTION_MASK_ISA_3DNOW_A

#define OPTION_MASK_ISA_SSE_UNSET \
  (OPTION_MASK_ISA_SSE | OPTION_MASK_ISA_SSE2_UNSET)
#define OPTION_MASK_ISA_SSE2_UNSET \
  (OPTION_MASK_ISA_SSE2 | OPTION_MASK_ISA_SSE3_UNSET)
#define OPTION_MASK_ISA_SSE3_UNSET \
  (OPTION_MASK_ISA_SSE3 \
   | OPTION_MASK_ISA_SSSE3_UNSET \
   | OPTION_MASK_ISA_SSE4A_UNSET )
#define OPTION_MASK_ISA_SSSE3_UNSET \
  (OPTION_MASK_ISA_SSSE3 | OPTION_MASK_ISA_SSE4_1_UNSET)
#define OPTION_MASK_ISA_SSE4_1_UNSET \
  (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSE4_2_UNSET)
#define OPTION_MASK_ISA_SSE4_2_UNSET \
  (OPTION_MASK_ISA_SSE4_2 | OPTION_MASK_ISA_AVX_UNSET )
#define OPTION_MASK_ISA_AVX_UNSET \
  (OPTION_MASK_ISA_AVX | OPTION_MASK_ISA_FMA_UNSET \
   | OPTION_MASK_ISA_FMA4_UNSET)
#define OPTION_MASK_ISA_FMA_UNSET OPTION_MASK_ISA_FMA

/* SSE4 includes both SSE4.1 and SSE4.2.  -mno-sse4 should the same
   as -mno-sse4.1. */
#define OPTION_MASK_ISA_SSE4_UNSET OPTION_MASK_ISA_SSE4_1_UNSET

#define OPTION_MASK_ISA_SSE4A_UNSET \
  (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_FMA4_UNSET)

#define OPTION_MASK_ISA_FMA4_UNSET \
  (OPTION_MASK_ISA_FMA4 | OPTION_MASK_ISA_XOP_UNSET)
#define OPTION_MASK_ISA_XOP_UNSET OPTION_MASK_ISA_XOP
#define OPTION_MASK_ISA_LWP_UNSET OPTION_MASK_ISA_LWP

#define OPTION_MASK_ISA_AES_UNSET OPTION_MASK_ISA_AES
#define OPTION_MASK_ISA_PCLMUL_UNSET OPTION_MASK_ISA_PCLMUL
#define OPTION_MASK_ISA_ABM_UNSET OPTION_MASK_ISA_ABM
#define OPTION_MASK_ISA_POPCNT_UNSET OPTION_MASK_ISA_POPCNT
#define OPTION_MASK_ISA_CX16_UNSET OPTION_MASK_ISA_CX16
#define OPTION_MASK_ISA_SAHF_UNSET OPTION_MASK_ISA_SAHF
#define OPTION_MASK_ISA_MOVBE_UNSET OPTION_MASK_ISA_MOVBE
#define OPTION_MASK_ISA_CRC32_UNSET OPTION_MASK_ISA_CRC32

#if 0
/* Vectorization library interface and handlers.  */
tree (*ix86_veclib_handler)(enum built_in_function, tree, tree) = NULL;
static tree ix86_veclibabi_svml (enum built_in_function, tree, tree);
static tree ix86_veclibabi_acml (enum built_in_function, tree, tree);
#endif

/* Processor target table, indexed by processor number */
struct ptt
{
  const struct processor_costs *cost;		/* Processor costs */
  const int align_loop;				/* Default alignments.  */
  const int align_loop_max_skip;
  const int align_jump;
  const int align_jump_max_skip;
  const int align_func;
};


/* Return the "natural" mode for TYPE.  In most cases, this is just TYPE_MODE.
   But in the case of vector types, it is some vector mode.

   When we have only some of our vector isa extensions enabled, then there
   are some modes for which vector_mode_supported_p is false.  For these
   modes, the generic vector support in gcc will choose some non-vector mode
   in order to implement the type.  By computing the natural mode, we'll
   select the proper ABI location for the operand and not depend on whatever
   the middle-end decides to do with these vector types.

   The midde-end can't deal with the vector types > 16 bytes.  In this
   case, we return the original mode and warn ABI change if CUM isn't
   NULL.  */

enum machine_mode
type_natural_mode (const_tree type, CUMULATIVE_ARGS *cum)
{
  enum machine_mode mode = TYPE_MODE (type);

  if (TREE_CODE (type) == VECTOR_TYPE && !VECTOR_MODE_P (mode))
    {
      HOST_WIDE_INT size = int_size_in_bytes (type);
      if ((size == 8 || size == 16 || size == 32)
	  /* ??? Generic code allows us to create width 1 vectors.  Ignore.  */
	  && TYPE_VECTOR_SUBPARTS (type) > 1)
	{
	  enum machine_mode innermode = TYPE_MODE (TREE_TYPE (type));

	  if (TREE_CODE (TREE_TYPE (type)) == REAL_TYPE)
	    mode = MIN_MODE_VECTOR_FLOAT;
	  else
	    mode = MIN_MODE_VECTOR_INT;

	  /* Get the mode which has this inner mode and number of units.  */
	  for (; mode != VOIDmode; mode = GET_MODE_WIDER_MODE (mode))
	    if (GET_MODE_NUNITS (mode) == TYPE_VECTOR_SUBPARTS (type)
		&& GET_MODE_INNER (mode) == innermode)
	      {
		if (size == 32 && !TARGET_AVX)
		  {
		    static bool warnedavx;

		    if (cum
			&& !warnedavx 
			&& cum->warn_avx)
		      {
			warnedavx = true;
			warning (0, "AVX vector argument without AVX "
				 "enabled changes the ABI");
		      }
		    return TYPE_MODE (type);
		  }
		else
		  return mode;
	      }

	  gcc_unreachable ();
	}
    }

  return mode;
}

/* x86-64 register passing implementation.  See x86-64 ABI for details.  Goal
   of this code is to classify each 8bytes of incoming argument by the register
   class and assign registers accordingly.  */

/* Return the union class of CLASS1 and CLASS2.
   See the x86-64 PS ABI for details.  */

static enum x86_64_reg_class
merge_classes (enum x86_64_reg_class class1, enum x86_64_reg_class class2)
{
  /* Rule #1: If both classes are equal, this is the resulting class.  */
  if (class1 == class2)
    return class1;

  /* Rule #2: If one of the classes is NO_CLASS, the resulting class is
     the other class.  */
  if (class1 == X86_64_NO_CLASS)
    return class2;
  if (class2 == X86_64_NO_CLASS)
    return class1;

  /* Rule #3: If one of the classes is MEMORY, the result is MEMORY.  */
  if (class1 == X86_64_MEMORY_CLASS || class2 == X86_64_MEMORY_CLASS)
    return X86_64_MEMORY_CLASS;

  /* Rule #4: If one of the classes is INTEGER, the result is INTEGER.  */
  if ((class1 == X86_64_INTEGERSI_CLASS && class2 == X86_64_SSESF_CLASS)
      || (class2 == X86_64_INTEGERSI_CLASS && class1 == X86_64_SSESF_CLASS))
    return X86_64_INTEGERSI_CLASS;
  if (class1 == X86_64_INTEGER_CLASS || class1 == X86_64_INTEGERSI_CLASS
      || class2 == X86_64_INTEGER_CLASS || class2 == X86_64_INTEGERSI_CLASS)
    return X86_64_INTEGER_CLASS;

  /* Rule #5: If one of the classes is X87, X87UP, or COMPLEX_X87 class,
     MEMORY is used.  */
  if (class1 == X86_64_X87_CLASS
      || class1 == X86_64_X87UP_CLASS
      || class1 == X86_64_COMPLEX_X87_CLASS
      || class2 == X86_64_X87_CLASS
      || class2 == X86_64_X87UP_CLASS
      || class2 == X86_64_COMPLEX_X87_CLASS)
    return X86_64_MEMORY_CLASS;

  /* Rule #6: Otherwise class SSE is used.  */
  return X86_64_SSE_CLASS;
}

/* Classify the argument of type TYPE and mode MODE.
   CLASSES will be filled by the register class used to pass each word
   of the operand.  The number of words is returned.  In case the parameter
   should be passed in memory, 0 is returned. As a special case for zero
   sized containers, classes[0] will be NO_CLASS and 1 is returned.

   BIT_OFFSET is used internally for handling records and specifies offset
   of the offset in bits modulo 256 to avoid overflow cases.

   See the x86-64 PS ABI for details.
*/

int
classify_argument (enum machine_mode mode, const_tree type,
		   enum x86_64_reg_class classes[MAX_CLASSES], int bit_offset)
{
  HOST_WIDE_INT bytes =
    (mode == BLKmode) ? int_size_in_bytes (type) : (int) GET_MODE_SIZE (mode);
  int words = (bytes + (bit_offset % 64) / 8 + UNITS_PER_WORD - 1) / UNITS_PER_WORD;

  /* Variable sized entities are always passed/returned in memory.  */
  if (bytes < 0)
    return 0;

  if (mode != VOIDmode
      && targetm.calls.must_pass_in_stack (mode, type))
    return 0;

  if (type && AGGREGATE_TYPE_P (type))
    {
      int i;
      tree field;
      enum x86_64_reg_class subclasses[MAX_CLASSES];

      /* On x86-64 we pass structures larger than 32 bytes on the stack.  */
      if (bytes > 32)
	return 0;

      for (i = 0; i < words; i++)
	classes[i] = X86_64_NO_CLASS;

      /* Zero sized arrays or structures are NO_CLASS.  We return 0 to
	 signalize memory class, so handle it as special case.  */
      if (!words)
	{
	  classes[0] = X86_64_NO_CLASS;
	  return 1;
	}

      /* Classify each field of record and merge classes.  */
      switch (TREE_CODE (type))
	{
	case RECORD_TYPE:
	  /* And now merge the fields of structure.  */
	  for (field = TYPE_FIELDS (type); field; field = TREE_CHAIN (field))
	    {
	      if (TREE_CODE (field) == FIELD_DECL)
		{
		  int num;

		  if (TREE_TYPE (field) == error_mark_node)
		    continue;

		  /* Bitfields are always classified as integer.  Handle them
		     early, since later code would consider them to be
		     misaligned integers.  */
		  if (DECL_BIT_FIELD (field))
		    {
		      for (i = (int_bit_position (field) + (bit_offset % 64)) / 8 / 8;
			   i < ((int_bit_position (field) + (bit_offset % 64))
			        + tree_low_cst (DECL_SIZE (field), 0)
				+ 63) / 8 / 8; i++)
			classes[i] =
			  merge_classes (X86_64_INTEGER_CLASS,
					 classes[i]);
		    }
		  else
		    {
		      int pos;

		      type = TREE_TYPE (field);

		      /* Flexible array member is ignored.  */
		      if (TYPE_MODE (type) == BLKmode
			  && TREE_CODE (type) == ARRAY_TYPE
			  && TYPE_SIZE (type) == NULL_TREE
			  && TYPE_DOMAIN (type) != NULL_TREE
			  && (TYPE_MAX_VALUE (TYPE_DOMAIN (type))
			      == NULL_TREE))
			{
			  static bool warned;
			  
			  if (!warned && warn_psabi)
			    {
			      warned = true;
			      inform (input_location,
				      "The ABI of passing struct with"
				      " a flexible array member has"
				      " changed in GCC 4.4");
			    }
			  continue;
			}
		      num = classify_argument (TYPE_MODE (type), type,
					       subclasses,
					       (int_bit_position (field)
						+ bit_offset) % 256);
		      if (!num)
			return 0;
		      pos = (int_bit_position (field) + (bit_offset % 64)) / 8 / 8;
		      for (i = 0; i < num && (i + pos) < words; i++)
			classes[i + pos] =
			  merge_classes (subclasses[i], classes[i + pos]);
		    }
		}
	    }
	  break;

	case ARRAY_TYPE:
	  /* Arrays are handled as small records.  */
	  {
	    int num;
	    num = classify_argument (TYPE_MODE (TREE_TYPE (type)),
				     TREE_TYPE (type), subclasses, bit_offset);
	    if (!num)
	      return 0;

	    /* The partial classes are now full classes.  */
	    if (subclasses[0] == X86_64_SSESF_CLASS && bytes != 4)
	      subclasses[0] = X86_64_SSE_CLASS;
	    if (subclasses[0] == X86_64_INTEGERSI_CLASS
		&& !((bit_offset % 64) == 0 && bytes == 4))
	      subclasses[0] = X86_64_INTEGER_CLASS;

	    for (i = 0; i < words; i++)
	      classes[i] = subclasses[i % num];

	    break;
	  }
	case UNION_TYPE:
	case QUAL_UNION_TYPE:
	  /* Unions are similar to RECORD_TYPE but offset is always 0.
	     */
	  for (field = TYPE_FIELDS (type); field; field = TREE_CHAIN (field))
	    {
	      if (TREE_CODE (field) == FIELD_DECL)
		{
		  int num;

		  if (TREE_TYPE (field) == error_mark_node)
		    continue;

		  num = classify_argument (TYPE_MODE (TREE_TYPE (field)),
					   TREE_TYPE (field), subclasses,
					   bit_offset);
		  if (!num)
		    return 0;
		  for (i = 0; i < num; i++)
		    classes[i] = merge_classes (subclasses[i], classes[i]);
		}
	    }
	  break;

	default:
	  gcc_unreachable ();
	}

      if (words > 2)
	{
	  /* When size > 16 bytes, if the first one isn't
	     X86_64_SSE_CLASS or any other ones aren't
	     X86_64_SSEUP_CLASS, everything should be passed in
	     memory.  */
	  if (classes[0] != X86_64_SSE_CLASS)
	      return 0;

	  for (i = 1; i < words; i++)
	    if (classes[i] != X86_64_SSEUP_CLASS)
	      return 0;
	}

      /* Final merger cleanup.  */
      for (i = 0; i < words; i++)
	{
	  /* If one class is MEMORY, everything should be passed in
	     memory.  */
	  if (classes[i] == X86_64_MEMORY_CLASS)
	    return 0;

	  /* The X86_64_SSEUP_CLASS should be always preceded by
	     X86_64_SSE_CLASS or X86_64_SSEUP_CLASS.  */
	  if (classes[i] == X86_64_SSEUP_CLASS
	      && classes[i - 1] != X86_64_SSE_CLASS
	      && classes[i - 1] != X86_64_SSEUP_CLASS)
	    {
	      /* The first one should never be X86_64_SSEUP_CLASS.  */
	      gcc_assert (i != 0);
	      classes[i] = X86_64_SSE_CLASS;
	    }

	  /*  If X86_64_X87UP_CLASS isn't preceded by X86_64_X87_CLASS,
	       everything should be passed in memory.  */
	  if (classes[i] == X86_64_X87UP_CLASS
	      && (classes[i - 1] != X86_64_X87_CLASS))
	    {
	      static bool warned;

	      /* The first one should never be X86_64_X87UP_CLASS.  */
	      gcc_assert (i != 0);
	      if (!warned && warn_psabi)
		{
		  warned = true;
		  inform (input_location,
			  "The ABI of passing union with long double"
			  " has changed in GCC 4.4");
		}
	      return 0;
	    }
	}
      return words;
    }

  /* Compute alignment needed.  We align all types to natural boundaries with
     exception of XFmode that is aligned to 64bits.  */
  if (mode != VOIDmode && mode != BLKmode)
    {
      int mode_alignment = GET_MODE_BITSIZE (mode);

      if (mode == XFmode)
	mode_alignment = 128;
      else if (mode == XCmode)
	mode_alignment = 256;
      if (COMPLEX_MODE_P (mode))
	mode_alignment /= 2;
      /* Misaligned fields are always returned in memory.  */
      if (bit_offset % mode_alignment)
	return 0;
    }

  /* for V1xx modes, just use the base mode */
  if (VECTOR_MODE_P (mode) && mode != V1DImode && mode != V1TImode
      && GET_MODE_SIZE (GET_MODE_INNER (mode)) == bytes)
    mode = GET_MODE_INNER (mode);

  /* Classification of atomic types.  */
  switch (mode)
    {
    case SDmode:
    case DDmode:
      classes[0] = X86_64_SSE_CLASS;
      return 1;
    case TDmode:
      classes[0] = X86_64_SSE_CLASS;
      classes[1] = X86_64_SSEUP_CLASS;
      return 2;
    case DImode:
    case SImode:
    case HImode:
    case QImode:
    case CSImode:
    case CHImode:
    case CQImode:
      {
	int size = (bit_offset % 64)+ (int) GET_MODE_BITSIZE (mode);

	if (size <= 32)
	  {
	    classes[0] = X86_64_INTEGERSI_CLASS;
	    return 1;
	  }
	else if (size <= 64)
	  {
	    classes[0] = X86_64_INTEGER_CLASS;
	    return 1;
	  }
	else if (size <= 64+32)
	  {
	    classes[0] = X86_64_INTEGER_CLASS;
	    classes[1] = X86_64_INTEGERSI_CLASS;
	    return 2;
	  }
	else if (size <= 64+64)
	  {
	    classes[0] = classes[1] = X86_64_INTEGER_CLASS;
	    return 2;
	  }
	else
	  gcc_unreachable ();
      }
    case CDImode:
    case TImode:
      classes[0] = classes[1] = X86_64_INTEGER_CLASS;
      return 2;
    case COImode:
    case OImode:
      /* OImode shouldn't be used directly.  */
      gcc_unreachable ();
    case CTImode:
      return 0;
    case SFmode:
      if (!(bit_offset % 64))
	classes[0] = X86_64_SSESF_CLASS;
      else
	classes[0] = X86_64_SSE_CLASS;
      return 1;
    case DFmode:
      classes[0] = X86_64_SSEDF_CLASS;
      return 1;
    case XFmode:
      classes[0] = X86_64_X87_CLASS;
      classes[1] = X86_64_X87UP_CLASS;
      return 2;
    case TFmode:
      classes[0] = X86_64_SSE_CLASS;
      classes[1] = X86_64_SSEUP_CLASS;
      return 2;
    case SCmode:
      classes[0] = X86_64_SSE_CLASS;
      if (!(bit_offset % 64))
	return 1;
      else
	{
	  static bool warned;

	  if (!warned && warn_psabi)
	    {
	      warned = true;
	      inform (input_location,
		      "The ABI of passing structure with complex float"
		      " member has changed in GCC 4.4");
	    }
	  classes[1] = X86_64_SSESF_CLASS;
	  return 2;
	}
    case DCmode:
      classes[0] = X86_64_SSEDF_CLASS;
      classes[1] = X86_64_SSEDF_CLASS;
      return 2;
    case XCmode:
      classes[0] = X86_64_COMPLEX_X87_CLASS;
      return 1;
    case TCmode:
      /* This modes is larger than 16 bytes.  */
      return 0;
    case V8SFmode:
    case V8SImode:
    case V32QImode:
    case V16HImode:
    case V4DFmode:
    case V4DImode:
      classes[0] = X86_64_SSE_CLASS;
      classes[1] = X86_64_SSEUP_CLASS;
      classes[2] = X86_64_SSEUP_CLASS;
      classes[3] = X86_64_SSEUP_CLASS;
      return 4;
    case V4SFmode:
    case V4SImode:
    case V16QImode:
    case V8HImode:
    case V2DFmode:
    case V2DImode:
      classes[0] = X86_64_SSE_CLASS;
      classes[1] = X86_64_SSEUP_CLASS;
      return 2;
    case V1TImode:
    case V1DImode:
    case V2SFmode:
    case V2SImode:
    case V4HImode:
    case V8QImode:
      classes[0] = X86_64_SSE_CLASS;
      return 1;
    case BLKmode:
    case VOIDmode:
      return 0;
    default:
      gcc_assert (VECTOR_MODE_P (mode));

      if (bytes > 16)
	return 0;

      gcc_assert (GET_MODE_CLASS (GET_MODE_INNER (mode)) == MODE_INT);

      if (bit_offset + GET_MODE_BITSIZE (mode) <= 32)
	classes[0] = X86_64_INTEGERSI_CLASS;
      else
	classes[0] = X86_64_INTEGER_CLASS;
      classes[1] = X86_64_INTEGER_CLASS;
      return 1 + (bytes > 8);
    }
}

/* Examine the argument and return set number of register required in each
   class.  Return 0 iff parameter should be passed in memory.  */
int
examine_argument (enum machine_mode mode, const_tree type, int in_return,
		  int *int_nregs, int *sse_nregs)
{
  enum x86_64_reg_class regclass[MAX_CLASSES];
  int n = classify_argument (mode, type, regclass, 0);

  *int_nregs = 0;
  *sse_nregs = 0;
  if (!n)
    return 0;
  for (n--; n >= 0; n--)
    switch (regclass[n])
      {
      case X86_64_INTEGER_CLASS:
      case X86_64_INTEGERSI_CLASS:
	(*int_nregs)++;
	break;
      case X86_64_SSE_CLASS:
      case X86_64_SSESF_CLASS:
      case X86_64_SSEDF_CLASS:
	(*sse_nregs)++;
	break;
      case X86_64_NO_CLASS:
      case X86_64_SSEUP_CLASS:
	break;
      case X86_64_X87_CLASS:
      case X86_64_X87UP_CLASS:
	if (!in_return)
	  return 0;
	break;
      case X86_64_COMPLEX_X87_CLASS:
	return in_return ? 2 : 0;
      case X86_64_MEMORY_CLASS:
	gcc_unreachable ();
      }
  return 1;
}

/* Return true when TYPE should be 128bit aligned for 32bit argument passing
   ABI.  */
bool
contains_aligned_value_p (tree type)
{
  enum machine_mode mode = TYPE_MODE (type);
  if (((TARGET_SSE && SSE_REG_MODE_P (mode))
       || mode == TDmode
       || mode == TFmode
       || mode == TCmode)
      && (!TYPE_USER_ALIGN (type) || TYPE_ALIGN (type) > 128))
    return true;
  if (TYPE_ALIGN (type) < 128)
    return false;

  if (AGGREGATE_TYPE_P (type))
    {
      /* Walk the aggregates recursively.  */
      switch (TREE_CODE (type))
	{
	case RECORD_TYPE:
	case UNION_TYPE:
	case QUAL_UNION_TYPE:
	  {
	    tree field;

	    /* Walk all the structure fields.  */
	    for (field = TYPE_FIELDS (type); field; field = TREE_CHAIN (field))
	      {
		if (TREE_CODE (field) == FIELD_DECL
		    && contains_aligned_value_p (TREE_TYPE (field)))
		  return true;
	      }
	    break;
	  }

	case ARRAY_TYPE:
	  /* Just for use if some languages passes arrays by value.  */
	  if (contains_aligned_value_p (TREE_TYPE (type)))
	    return true;
	  break;

	default:
	  gcc_unreachable ();
	}
    }
  return false;
}