sysdeps/x86_64/multiarch/init-arch.h

   1 /* This file is part of the GNU C Library.
   2    Copyright (C) 2008-2013 Free Software Foundation, Inc.
   3
   4    The GNU C Library is free software; you can redistribute it and/or
   5    modify it under the terms of the GNU Lesser General Public
   6    License as published by the Free Software Foundation; either
   7    version 2.1 of the License, or (at your option) any later version.
   8
   9    The GNU C Library is distributed in the hope that it will be useful,
  10    but WITHOUT ANY WARRANTY; without even the implied warranty of
  11    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12    Lesser General Public License for more details.
  13
  14    You should have received a copy of the GNU Lesser General Public
  15    License along with the GNU C Library; if not, see
  16    <http://www.gnu.org/licenses/>.  */
  17
  18 #define bit_Fast_Rep_String             (1 << 0)
  19 #define bit_Fast_Copy_Backward          (1 << 1)
  20 #define bit_Slow_BSF                    (1 << 2)
  21 #define bit_Prefer_SSE_for_memop        (1 << 3)
  22 #define bit_Fast_Unaligned_Load         (1 << 4)
  23 #define bit_Prefer_PMINUB_for_stringop  (1 << 5)
  24 #define bit_AVX_Usable                  (1 << 6)
  25 #define bit_FMA_Usable                  (1 << 7)
  26 #define bit_FMA4_Usable                 (1 << 8)
  27
  28 /* CPUID Feature flags.  */
  29 #define bit_SSE2        (1 << 26)
  30 #define bit_SSSE3       (1 << 9)
  31 #define bit_SSE4_1      (1 << 19)
  32 #define bit_SSE4_2      (1 << 20)
  33 #define bit_OSXSAVE     (1 << 27)
  34 #define bit_AVX         (1 << 28)
  35 #define bit_POPCOUNT    (1 << 23)
  36 #define bit_FMA         (1 << 12)
  37 #define bit_FMA4        (1 << 16)
  38
  39 /* XCR0 Feature flags.  */
  40 #define bit_XMM_state  (1 << 1)
  41 #define bit_YMM_state  (2 << 1)
  42
  43 #ifdef  __ASSEMBLER__
  44
  45 # include <ifunc-defines.h>
  46
  47 # define index_SSE2     COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
  48 # define index_SSSE3    COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
  49 # define index_SSE4_1   COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
  50 # define index_SSE4_2   COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
  51 # define index_AVX      COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
  52
  53 # define index_Fast_Rep_String          FEATURE_INDEX_1*FEATURE_SIZE
  54 # define index_Fast_Copy_Backward       FEATURE_INDEX_1*FEATURE_SIZE
  55 # define index_Slow_BSF                 FEATURE_INDEX_1*FEATURE_SIZE
  56 # define index_Prefer_SSE_for_memop     FEATURE_INDEX_1*FEATURE_SIZE
  57 # define index_Fast_Unaligned_Load      FEATURE_INDEX_1*FEATURE_SIZE
  58 # define index_Prefer_PMINUB_for_stringop FEATURE_INDEX_1*FEATURE_SIZE
  59 # define index_AVX_Usable               FEATURE_INDEX_1*FEATURE_SIZE
  60 # define index_FMA_Usable               FEATURE_INDEX_1*FEATURE_SIZE
  61 # define index_FMA4_Usable              FEATURE_INDEX_1*FEATURE_SIZE
  62
  63 #else   /* __ASSEMBLER__ */
  64
  65 # include <sys/param.h>
  66
  67 enum
  68   {
  69     COMMON_CPUID_INDEX_1 = 0,
  70     COMMON_CPUID_INDEX_80000001,        /* for AMD */
  71     /* Keep the following line at the end.  */
  72     COMMON_CPUID_INDEX_MAX
  73   };
  74
  75 enum
  76   {
  77     FEATURE_INDEX_1 = 0,
  78     /* Keep the following line at the end.  */
  79     FEATURE_INDEX_MAX
  80   };
  81
  82 extern struct cpu_features
  83 {
  84   enum cpu_features_kind
  85     {
  86       arch_kind_unknown = 0,
  87       arch_kind_intel,
  88       arch_kind_amd,
  89       arch_kind_other
  90     } kind;
  91   int max_cpuid;
  92   struct cpuid_registers
  93   {
  94     unsigned int eax;
  95     unsigned int ebx;
  96     unsigned int ecx;
  97     unsigned int edx;
  98   } cpuid[COMMON_CPUID_INDEX_MAX];
  99   unsigned int family;
 100   unsigned int model;
 101   unsigned int feature[FEATURE_INDEX_MAX];
 102 } __cpu_features attribute_hidden;
 103
 104
 105 extern void __init_cpu_features (void) attribute_hidden;
 106 # define INIT_ARCH() \
 107   do                                                    \
 108     if (__cpu_features.kind == arch_kind_unknown)       \
 109       __init_cpu_features ();                           \
 110   while (0)
 111
 112 /* Used from outside libc.so to get access to the CPU features structure.  */
 113 extern const struct cpu_features *__get_cpu_features (void)
 114      __attribute__ ((const));
 115
 116 # ifndef NOT_IN_libc
 117 #  define __get_cpu_features()  (&__cpu_features)
 118 # endif
 119
 120 # define HAS_CPU_FEATURE(idx, reg, bit) \
 121   ((__get_cpu_features ()->cpuid[idx].reg & (bit)) != 0)
 122
 123 /* Following are the feature tests used throughout libc.  */
 124
 125 /* CPUID_* evaluates to true if the feature flag is enabled.
 126    We always use &__cpu_features because the HAS_CPUID_* macros
 127    are called only within __init_cpu_features, where we can't
 128    call __get_cpu_features without infinite recursion.  */
 129 # define HAS_CPUID_FLAG(idx, reg, bit) \
 130   (((&__cpu_features)->cpuid[idx].reg & (bit)) != 0)
 131
 132 # define CPUID_OSXSAVE \
 133   HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_OSXSAVE)
 134 # define CPUID_AVX \
 135   HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_AVX)
 136 # define CPUID_FMA \
 137   HAS_CPUID_FLAG (COMMON_CPUID_INDEX_1, ecx, bit_FMA)
 138 # define CPUID_FMA4 \
 139   HAS_CPUID_FLAG (COMMON_CPUID_INDEX_80000001, ecx, bit_FMA4)
 140
 141 /* HAS_* evaluates to true if we may use the feature at runtime.  */
 142 # define HAS_SSE2       HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, edx, bit_SSE2)
 143 # define HAS_POPCOUNT   HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_POPCOUNT)
 144 # define HAS_SSSE3      HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSSE3)
 145 # define HAS_SSE4_1     HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_1)
 146 # define HAS_SSE4_2     HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_2)
 147
 148 # define index_Fast_Rep_String          FEATURE_INDEX_1
 149 # define index_Fast_Copy_Backward       FEATURE_INDEX_1
 150 # define index_Slow_BSF                 FEATURE_INDEX_1
 151 # define index_Prefer_SSE_for_memop     FEATURE_INDEX_1
 152 # define index_Fast_Unaligned_Load      FEATURE_INDEX_1
 153 # define index_AVX_Usable               FEATURE_INDEX_1
 154 # define index_FMA_Usable               FEATURE_INDEX_1
 155 # define index_FMA4_Usable              FEATURE_INDEX_1
 156
 157 # define HAS_ARCH_FEATURE(name) \
 158   ((__get_cpu_features ()->feature[index_##name] & (bit_##name)) != 0)
 159
 160 # define HAS_FAST_REP_STRING            HAS_ARCH_FEATURE (Fast_Rep_String)
 161 # define HAS_FAST_COPY_BACKWARD         HAS_ARCH_FEATURE (Fast_Copy_Backward)
 162 # define HAS_SLOW_BSF                   HAS_ARCH_FEATURE (Slow_BSF)
 163 # define HAS_PREFER_SSE_FOR_MEMOP       HAS_ARCH_FEATURE (Prefer_SSE_for_memop)
 164 # define HAS_FAST_UNALIGNED_LOAD        HAS_ARCH_FEATURE (Fast_Unaligned_Load)
 165 # define HAS_AVX                        HAS_ARCH_FEATURE (AVX_Usable)
 166 # define HAS_FMA                        HAS_ARCH_FEATURE (FMA_Usable)
 167 # define HAS_FMA4                       HAS_ARCH_FEATURE (FMA4_Usable)
 168
 169 #endif  /* __ASSEMBLER__ */