eng_devcrypto: make sure digest can do copy
[openssl.git] / crypto / ia64cpuid.S
index dd27e163ce03d162cd51174e7fb6ee7dac546efc..92c55124bd374229e46119f85f59dd5c7e4675a9 100644 (file)
@@ -1,7 +1,19 @@
+// Copyright 2004-2017 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the OpenSSL license (the "License").  You may not use
+// this file except in compliance with the License.  You can obtain a copy
+// in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
 // Works on all IA-64 platforms: Linux, HP-UX, Win64i...
 // On Win64i compile with ias.exe.
 .text
 
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+#define        ADDP    addp4
+#else
+#define        ADDP    add
+#endif
+
 .global        OPENSSL_cpuid_setup#
 .proc  OPENSSL_cpuid_setup#
 OPENSSL_cpuid_setup:
@@ -63,7 +75,7 @@ OPENSSL_wipe_cpu:
 { .mii;        add             r9=96*8-8,r9
        mov             ar.ec=1         };;
 
-// One can sweep double as fast, but then we can't quarantee
+// One can sweep double as fast, but then we can't guarantee
 // that backing storage is wiped...
 .L_wipe_top:
 { .mfi;        st8             [r9]=r0,-8
@@ -131,9 +143,7 @@ OPENSSL_wipe_cpu:
 .proc  OPENSSL_cleanse#
 OPENSSL_cleanse:
 { .mib;        cmp.eq          p6,p0=0,r33         // len==0
-#if defined(_HPUX_SOURCE) && !defined(_LP64)
-       addp4           r32=0,r32
-#endif
+       ADDP            r32=0,r32
 (p6)   br.ret.spnt     b0              };;
 { .mib;        and             r2=7,r32
        cmp.leu         p6,p0=15,r33        // len>=15
@@ -166,14 +176,51 @@ OPENSSL_cleanse:
 (p6)   br.ret.sptk.many        b0      };;
 .endp  OPENSSL_cleanse#
 
+.global        CRYPTO_memcmp#
+.proc  CRYPTO_memcmp#
+.align 32
+.skip  16
+CRYPTO_memcmp:
+       .prologue
+{ .mib;        mov             r8=0
+       cmp.eq          p6,p0=0,r34         // len==0?
+(p6)   br.ret.spnt     b0              };;
+       .save           ar.pfs,r2
+{ .mib;        alloc           r2=ar.pfs,3,5,0,8
+       .save           ar.lc,r3
+       mov             r3=ar.lc
+       brp.loop.imp    .Loop_cmp_ctop,.Loop_cmp_cend-16
+                                       }
+{ .mib;        sub             r10=r34,r0,1
+       .save           pr,r9
+       mov             r9=pr           };;
+{ .mii;        ADDP            r16=0,r32
+       mov             ar.lc=r10
+       mov             ar.ec=4         }
+{ .mib;        ADDP            r17=0,r33
+       mov             pr.rot=1<<16    };;
+
+.Loop_cmp_ctop:
+{ .mib;        (p16)   ld1     r32=[r16],1
+       (p18)   xor     r34=r34,r38     }
+{ .mib;        (p16)   ld1     r36=[r17],1
+       (p19)   or      r8=r8,r35
+       br.ctop.sptk    .Loop_cmp_ctop  };;
+.Loop_cmp_cend:
+
+{ .mib;        cmp.ne          p6,p0=0,r8
+       mov             ar.lc=r3        };;
+{ .mib;
+(p6)   mov             r8=1
+       mov             pr=r9,0x1ffff
+       br.ret.sptk.many        b0      };;
+.endp  CRYPTO_memcmp#
+
 .global        OPENSSL_instrument_bus#
 .proc  OPENSSL_instrument_bus#
-OPENSSL_instrument_cache:
+OPENSSL_instrument_bus:
 { .mmi;        mov             r2=r33
-#if defined(_HPUX_SOURCE) && !defined(_LP64)
-       addp4           r32=0,r32
-#endif
-                                       }
+       ADDP            r32=0,r32       }
 { .mmi;        mov             r8=ar.itc;;
        mov             r10=r0
        mov             r9=r8           };;
@@ -206,12 +253,9 @@ OPENSSL_instrument_cache:
 
 .global        OPENSSL_instrument_bus2#
 .proc  OPENSSL_instrument_bus2#
-OPENSSL_instrument_cache2:
+OPENSSL_instrument_bus2:
 { .mmi;        mov             r2=r33                  // put aside cnt
-#if defined(_HPUX_SOURCE) && !defined(_LP64)
-       addp4           r32=0,r32
-#endif
-                                       }
+       ADDP            r32=0,r32       }
 { .mmi;        mov             r8=ar.itc;;
        mov             r10=r0
        mov             r9=r8           };;