crypto/ia64cpuid.S

   1 // Works on all IA-64 platforms: Linux, HP-UX, Win64i...
   2 // On Win64i compile with ias.exe.
   3 .text
   4
   5 .global OPENSSL_cpuid_setup#
   6 .proc   OPENSSL_cpuid_setup#
   7 OPENSSL_cpuid_setup:
   8 { .mib; br.ret.sptk.many        b0              };;
   9 .endp   OPENSSL_cpuid_setup#
  10
  11 .global OPENSSL_rdtsc#
  12 .proc   OPENSSL_rdtsc#
  13 OPENSSL_rdtsc:
  14 { .mib; mov                     r8=ar.itc
  15         br.ret.sptk.many        b0              };;
  16 .endp   OPENSSL_rdtsc#
  17
  18 .global OPENSSL_atomic_add#
  19 .proc   OPENSSL_atomic_add#
  20 .align  32
  21 OPENSSL_atomic_add:
  22 { .mii; ld4             r2=[r32]
  23         nop.i           0
  24         nop.i           0               };;
  25 .Lspin:
  26 { .mii; mov             ar.ccv=r2
  27         add             r8=r2,r33
  28         mov             r3=r2           };;
  29 { .mmi; mf;;
  30         cmpxchg4.acq    r2=[r32],r8,ar.ccv
  31         nop.i           0               };;
  32 { .mib; cmp.ne          p6,p0=r2,r3
  33         nop.i           0
  34 (p6)    br.dpnt         .Lspin          };;
  35 { .mib; nop.m           0
  36         sxt4            r8=r8
  37         br.ret.sptk.many        b0      };;
  38 .endp   OPENSSL_atomic_add#
  39
  40 // Returns a structure comprising pointer to the top of stack of
  41 // the caller and pointer beyond backing storage for the current
  42 // register frame. The latter is required, because it might be
  43 // insufficient to wipe backing storage for the current frame
  44 // (as this procedure does), one might have to go further, toward
  45 // higher addresses to reach for whole "retroactively" saved
  46 // context...
  47 .global OPENSSL_wipe_cpu#
  48 .proc   OPENSSL_wipe_cpu#
  49 .align  32
  50 OPENSSL_wipe_cpu:
  51         .prologue
  52         .fframe 0
  53         .save   ar.pfs,r2
  54         .save   ar.lc,r3
  55 { .mib; alloc           r2=ar.pfs,0,96,0,96
  56         mov             r3=ar.lc
  57         brp.loop.imp    .L_wipe_top,.L_wipe_end-16
  58                                         };;
  59 { .mii; mov             r9=ar.bsp
  60         mov             r8=pr
  61         mov             ar.lc=96        };;
  62         .body
  63 { .mii; add             r9=96*8-8,r9
  64         mov             ar.ec=1         };;
  65
  66 // One can sweep double as fast, but then we can't quarantee
  67 // that backing storage is wiped...
  68 .L_wipe_top:
  69 { .mfi; st8             [r9]=r0,-8
  70         mov             f127=f0
  71         mov             r127=r0         }
  72 { .mfb; nop.m           0
  73         nop.f           0
  74         br.ctop.sptk    .L_wipe_top     };;
  75 .L_wipe_end:
  76
  77 { .mfi; mov             r11=r0
  78         mov             f6=f0
  79         mov             r14=r0          }
  80 { .mfi; mov             r15=r0
  81         mov             f7=f0
  82         mov             r16=r0          }
  83 { .mfi; mov             r17=r0
  84         mov             f8=f0
  85         mov             r18=r0          }
  86 { .mfi; mov             r19=r0
  87         mov             f9=f0
  88         mov             r20=r0          }
  89 { .mfi; mov             r21=r0
  90         mov             f10=f0
  91         mov             r22=r0          }
  92 { .mfi; mov             r23=r0
  93         mov             f11=f0
  94         mov             r24=r0          }
  95 { .mfi; mov             r25=r0
  96         mov             f12=f0
  97         mov             r26=r0          }
  98 { .mfi; mov             r27=r0
  99         mov             f13=f0
 100         mov             r28=r0          }
 101 { .mfi; mov             r29=r0
 102         mov             f14=f0
 103         mov             r30=r0          }
 104 { .mfi; mov             r31=r0
 105         mov             f15=f0
 106         nop.i           0               }
 107 { .mfi; mov             f16=f0          }
 108 { .mfi; mov             f17=f0          }
 109 { .mfi; mov             f18=f0          }
 110 { .mfi; mov             f19=f0          }
 111 { .mfi; mov             f20=f0          }
 112 { .mfi; mov             f21=f0          }
 113 { .mfi; mov             f22=f0          }
 114 { .mfi; mov             f23=f0          }
 115 { .mfi; mov             f24=f0          }
 116 { .mfi; mov             f25=f0          }
 117 { .mfi; mov             f26=f0          }
 118 { .mfi; mov             f27=f0          }
 119 { .mfi; mov             f28=f0          }
 120 { .mfi; mov             f29=f0          }
 121 { .mfi; mov             f30=f0          }
 122 { .mfi; add             r9=96*8+8,r9
 123         mov             f31=f0
 124         mov             pr=r8,0x1ffff   }
 125 { .mib; mov             r8=sp
 126         mov             ar.lc=r3
 127         br.ret.sptk     b0              };;
 128 .endp   OPENSSL_wipe_cpu#
 129
 130 .global OPENSSL_cleanse#
 131 .proc   OPENSSL_cleanse#
 132 OPENSSL_cleanse:
 133 { .mib; cmp.eq          p6,p0=0,r33         // len==0
 134 #if defined(_HPUX_SOURCE) && !defined(_LP64)
 135         addp4           r32=0,r32
 136 #endif
 137 (p6)    br.ret.spnt     b0              };;
 138 { .mib; and             r2=7,r32
 139         cmp.leu         p6,p0=15,r33        // len>=15
 140 (p6)    br.cond.dptk    .Lot            };;
 141
 142 .Little:
 143 { .mib; st1             [r32]=r0,1
 144         cmp.ltu         p6,p7=1,r33     }  // len>1
 145 { .mbb; add             r33=-1,r33         // len--
 146 (p6)    br.cond.dptk    .Little
 147 (p7)    br.ret.sptk.many        b0      };;
 148
 149 .Lot:
 150 { .mib; cmp.eq          p6,p0=0,r2
 151 (p6)    br.cond.dptk    .Laligned       };;
 152 { .mmi; st1             [r32]=r0,1;;
 153         and             r2=7,r32        }
 154 { .mib; add             r33=-1,r33
 155         br              .Lot            };;
 156
 157 .Laligned:
 158 { .mmi; st8             [r32]=r0,8
 159         and             r2=-8,r33           // len&~7
 160         add             r33=-8,r33      };; // len-=8
 161 { .mib; cmp.ltu         p6,p0=8,r2          // ((len+8)&~7)>8
 162 (p6)    br.cond.dptk    .Laligned       };;
 163
 164 { .mbb; cmp.eq          p6,p7=r0,r33
 165 (p7)    br.cond.dpnt    .Little
 166 (p6)    br.ret.sptk.many        b0      };;
 167 .endp   OPENSSL_cleanse#
 168
 169 .global OPENSSL_instrument_bus#
 170 .proc   OPENSSL_instrument_bus#
 171 OPENSSL_instrument_bus:
 172 { .mmi; mov             r2=r33
 173 #if defined(_HPUX_SOURCE) && !defined(_LP64)
 174         addp4           r32=0,r32
 175 #endif
 176                                         }
 177 { .mmi; mov             r8=ar.itc;;
 178         mov             r10=r0
 179         mov             r9=r8           };;
 180
 181 { .mmi; fc              r32;;
 182         ld4             r8=[r32]        };;
 183 { .mmi; mf
 184         mov             ar.ccv=r8
 185         add             r8=r8,r10       };;
 186 { .mmi; cmpxchg4.acq    r3=[r32],r8,ar.ccv
 187                                         };;
 188 .Loop:
 189 { .mmi; mov             r8=ar.itc;;
 190         sub             r10=r8,r9               // diff=tick-lasttick
 191         mov             r9=r8           };;     // lasttick=tick
 192 { .mmi; fc              r32;;
 193         ld4             r8=[r32]        };;
 194 { .mmi; mf
 195         mov             ar.ccv=r8
 196         add             r8=r8,r10       };;
 197 { .mmi; cmpxchg4.acq    r3=[r32],r8,ar.ccv
 198         add             r33=-1,r33
 199         add             r32=4,r32       };;
 200 { .mib; cmp4.ne         p6,p0=0,r33
 201 (p6)    br.cond.dptk    .Loop           };;
 202
 203 { .mib; sub             r8=r2,r33
 204         br.ret.sptk.many        b0      };;
 205 .endp   OPENSSL_instrument_bus#
 206
 207 .global OPENSSL_instrument_bus2#
 208 .proc   OPENSSL_instrument_bus2#
 209 OPENSSL_instrument_bus2:
 210 { .mmi; mov             r2=r33                  // put aside cnt
 211 #if defined(_HPUX_SOURCE) && !defined(_LP64)
 212         addp4           r32=0,r32
 213 #endif
 214                                         }
 215 { .mmi; mov             r8=ar.itc;;
 216         mov             r10=r0
 217         mov             r9=r8           };;
 218
 219 { .mmi; fc              r32;;
 220         ld4             r8=[r32]        };;
 221 { .mmi; mf
 222         mov             ar.ccv=r8
 223         add             r8=r8,r10       };;
 224 { .mmi; cmpxchg4.acq    r3=[r32],r8,ar.ccv
 225                                         };;
 226
 227 { .mmi; mov             r8=ar.itc;;
 228         sub             r10=r8,r9
 229         mov             r9=r8           };;
 230 .Loop2:
 231 { .mmi; mov             r11=r10                 // lastdiff=diff
 232         add             r34=-1,r34      };;     // --max
 233 { .mmi; fc              r32;;
 234         ld4             r8=[r32]
 235         cmp4.eq         p6,p0=0,r34     };;
 236 { .mmi; mf
 237         mov             ar.ccv=r8
 238         add             r8=r8,r10       };;
 239 { .mmb; cmpxchg4.acq    r3=[r32],r8,ar.ccv
 240 (p6)    br.cond.spnt    .Ldone2         };;
 241
 242 { .mmi; mov             r8=ar.itc;;
 243         sub             r10=r8,r9               // diff=tick-lasttick
 244         mov             r9=r8           };;     // lasttick=tick
 245 { .mmi; cmp.ne          p6,p0=r10,r11;;         // diff!=lastdiff
 246 (p6)    add             r33=-1,r33      };;     // conditional --cnt
 247 { .mib; cmp4.ne         p7,p0=0,r33
 248 (p6)    add             r32=4,r32               // conditional ++out
 249 (p7)    br.cond.dptk    .Loop2          };;
 250 .Ldone2:
 251 { .mib; sub             r8=r2,r33
 252         br.ret.sptk.many        b0      };;
 253 .endp   OPENSSL_instrument_bus2#