[openssl-commits] [openssl] master update

Andy Polyakov appro at openssl.org
Mon Jun 5 17:36:00 UTC 2017


The branch master has been updated
       via  71dd3b6464df1ba524601c3705e1d1e3c7c5406d (commit)
       via  22f9fa6e06305f85b45cf23fc68145d789633641 (commit)
       via  1ded2dd3ee9389b412e2ef0cf8e0b40a4ed3179b (commit)
       via  c83a4db52123603fd6d2fe8535dd19ec9e5d848d (commit)
      from  8fc063dcc9668589fd95533d25932396d60987f9 (commit)


- Log -----------------------------------------------------------------
commit 71dd3b6464df1ba524601c3705e1d1e3c7c5406d
Author: Andy Polyakov <appro at openssl.org>
Date:   Thu Jun 1 21:06:26 2017 +0200

    sha/keccak1600.c: add #ifdef KECCAK1600_ASM.
    
    Reviewed-by: Rich Salz <rsalz at openssl.org>

commit 22f9fa6e06305f85b45cf23fc68145d789633641
Author: Andy Polyakov <appro at openssl.org>
Date:   Mon Jun 5 11:07:55 2017 +0200

    sha/keccak1600.c: reduce temporary storage utilization even futher.
    
    Reviewed-by: Rich Salz <rsalz at openssl.org>

commit 1ded2dd3ee9389b412e2ef0cf8e0b40a4ed3179b
Author: Andy Polyakov <appro at openssl.org>
Date:   Sat Jun 3 22:07:05 2017 +0200

    sha/keccak1600.c: add another 1x variant.
    
    Reviewed-by: Rich Salz <rsalz at openssl.org>

commit c83a4db52123603fd6d2fe8535dd19ec9e5d848d
Author: Andy Polyakov <appro at openssl.org>
Date:   Wed May 31 12:12:44 2017 +0200

    sha/keccak1600.c: add ARM-specific "reference" tweaks.
    
    Reviewed-by: Rich Salz <rsalz at openssl.org>

-----------------------------------------------------------------------

Summary of changes:
 crypto/sha/keccak1600.c | 212 +++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 191 insertions(+), 21 deletions(-)

diff --git a/crypto/sha/keccak1600.c b/crypto/sha/keccak1600.c
index c89188f..2517bc6 100644
--- a/crypto/sha/keccak1600.c
+++ b/crypto/sha/keccak1600.c
@@ -11,6 +11,8 @@
 #include <string.h>
 #include <assert.h>
 
+#ifndef KECCAK1600_ASM
+
 #define ROL32(a, offset) (((a) << (offset)) | ((a) >> ((32 - (offset)) & 31)))
 
 static uint64_t ROL64(uint64_t val, int offset)
@@ -223,7 +225,8 @@ void KeccakF1600(uint64_t A[5][5])
  */
 static void Round(uint64_t A[5][5], size_t i)
 {
-    uint64_t C[5], D[5], T[2][5];
+    uint64_t C[5], E[2];        /* registers */
+    uint64_t D[5], T[2][5];     /* memory    */
 
     assert(i < (sizeof(iotas) / sizeof(iotas[0])));
 
@@ -233,41 +236,60 @@ static void Round(uint64_t A[5][5], size_t i)
     C[3] = A[0][3] ^ A[1][3] ^ A[2][3] ^ A[3][3] ^ A[4][3];
     C[4] = A[0][4] ^ A[1][4] ^ A[2][4] ^ A[3][4] ^ A[4][4];
 
+#if defined(__arm__)
+    D[1] = E[0] = ROL64(C[2], 1) ^ C[0];
+    D[4] = E[1] = ROL64(C[0], 1) ^ C[3];
+    D[0] = C[0] = ROL64(C[1], 1) ^ C[4];
+    D[2] = C[1] = ROL64(C[3], 1) ^ C[1];
+    D[3] = C[2] = ROL64(C[4], 1) ^ C[2];
+
+    T[0][0] = A[3][0] ^ C[0]; /* borrow T[0][0] */
+    T[0][1] = A[0][1] ^ E[0]; /* D[1] */
+    T[0][2] = A[0][2] ^ C[1]; /* D[2] */
+    T[0][3] = A[0][3] ^ C[2]; /* D[3] */
+    T[0][4] = A[0][4] ^ E[1]; /* D[4] */
+
+    C[3] = ROL64(A[3][3] ^ C[2], rhotates[3][3]);   /* D[3] */
+    C[4] = ROL64(A[4][4] ^ E[1], rhotates[4][4]);   /* D[4] */
+    C[0] =       A[0][0] ^ C[0]; /* rotate by 0 */  /* D[0] */
+    C[2] = ROL64(A[2][2] ^ C[1], rhotates[2][2]);   /* D[2] */
+    C[1] = ROL64(A[1][1] ^ E[0], rhotates[1][1]);   /* D[1] */
+#else
     D[0] = ROL64(C[1], 1) ^ C[4];
     D[1] = ROL64(C[2], 1) ^ C[0];
     D[2] = ROL64(C[3], 1) ^ C[1];
     D[3] = ROL64(C[4], 1) ^ C[2];
     D[4] = ROL64(C[0], 1) ^ C[3];
 
-    C[0] =       A[0][0] ^ D[0]; /* rotate by 0 */
-    C[1] = ROL64(A[1][1] ^ D[1], rhotates[1][1]);
-    C[2] = ROL64(A[2][2] ^ D[2], rhotates[2][2]);
-    C[3] = ROL64(A[3][3] ^ D[3], rhotates[3][3]);
-    C[4] = ROL64(A[4][4] ^ D[4], rhotates[4][4]);
-
     T[0][0] = A[3][0] ^ D[0]; /* borrow T[0][0] */
     T[0][1] = A[0][1] ^ D[1];
     T[0][2] = A[0][2] ^ D[2];
     T[0][3] = A[0][3] ^ D[3];
     T[0][4] = A[0][4] ^ D[4];
 
+    C[0] =       A[0][0] ^ D[0]; /* rotate by 0 */
+    C[1] = ROL64(A[1][1] ^ D[1], rhotates[1][1]);
+    C[2] = ROL64(A[2][2] ^ D[2], rhotates[2][2]);
+    C[3] = ROL64(A[3][3] ^ D[3], rhotates[3][3]);
+    C[4] = ROL64(A[4][4] ^ D[4], rhotates[4][4]);
+#endif
     A[0][0] = C[0] ^ (~C[1] & C[2]) ^ iotas[i];
     A[0][1] = C[1] ^ (~C[2] & C[3]);
     A[0][2] = C[2] ^ (~C[3] & C[4]);
     A[0][3] = C[3] ^ (~C[4] & C[0]);
     A[0][4] = C[4] ^ (~C[0] & C[1]);
 
-    C[0] = ROL64(T[0][3],        rhotates[0][3]);
-    C[1] = ROL64(A[1][4] ^ D[4], rhotates[1][4]);
-    C[2] = ROL64(A[2][0] ^ D[0], rhotates[2][0]);
-    C[3] = ROL64(A[3][1] ^ D[1], rhotates[3][1]);
-    C[4] = ROL64(A[4][2] ^ D[2], rhotates[4][2]);
+    T[1][0] = A[1][0] ^ (C[3] = D[0]);
+    T[1][1] = A[2][1] ^ (C[4] = D[1]); /* borrow T[1][1] */
+    T[1][2] = A[1][2] ^ (E[0] = D[2]);
+    T[1][3] = A[1][3] ^ (E[1] = D[3]);
+    T[1][4] = A[2][4] ^ (C[2] = D[4]); /* borrow T[1][4] */
 
-    T[1][0] = A[1][0] ^ D[0];
-    T[1][1] = A[2][1] ^ D[1]; /* borrow T[1][1] */
-    T[1][2] = A[1][2] ^ D[2];
-    T[1][3] = A[1][3] ^ D[3];
-    T[1][4] = A[2][4] ^ D[4]; /* borrow T[1][4] */
+    C[0] = ROL64(T[0][3],        rhotates[0][3]);
+    C[1] = ROL64(A[1][4] ^ C[2], rhotates[1][4]);   /* D[4] */
+    C[2] = ROL64(A[2][0] ^ C[3], rhotates[2][0]);   /* D[0] */
+    C[3] = ROL64(A[3][1] ^ C[4], rhotates[3][1]);   /* D[1] */
+    C[4] = ROL64(A[4][2] ^ E[0], rhotates[4][2]);   /* D[2] */
 
     A[1][0] = C[0] ^ (~C[1] & C[2]);
     A[1][1] = C[1] ^ (~C[2] & C[3]);
@@ -321,6 +343,149 @@ void KeccakF1600(uint64_t A[5][5])
     }
 }
 
+#elif defined(KECCAK_1X_ALT)
+/*
+ * This is variant of above KECCAK_1X that reduces requirement for
+ * temporary storage even further, but at cost of more updates to A[][].
+ * It's less suitable if A[][] is memory bound, but better if it's
+ * register bound.
+ */
+
+static void Round(uint64_t A[5][5], size_t i)
+{
+    uint64_t C[5], D[5];
+
+    assert(i < (sizeof(iotas) / sizeof(iotas[0])));
+
+    C[0] = A[0][0] ^ A[1][0] ^ A[2][0] ^ A[3][0] ^ A[4][0];
+    C[1] = A[0][1] ^ A[1][1] ^ A[2][1] ^ A[3][1] ^ A[4][1];
+    C[2] = A[0][2] ^ A[1][2] ^ A[2][2] ^ A[3][2] ^ A[4][2];
+    C[3] = A[0][3] ^ A[1][3] ^ A[2][3] ^ A[3][3] ^ A[4][3];
+    C[4] = A[0][4] ^ A[1][4] ^ A[2][4] ^ A[3][4] ^ A[4][4];
+
+    D[1] = C[0] ^  ROL64(C[2], 1);
+    D[2] = C[1] ^  ROL64(C[3], 1);
+    D[3] = C[2] ^= ROL64(C[4], 1);
+    D[4] = C[3] ^= ROL64(C[0], 1);
+    D[0] = C[4] ^= ROL64(C[1], 1);
+
+    A[0][1] ^= D[1];
+    A[1][1] ^= D[1];
+    A[2][1] ^= D[1];
+    A[3][1] ^= D[1];
+    A[4][1] ^= D[1];
+
+    A[0][2] ^= D[2];
+    A[1][2] ^= D[2];
+    A[2][2] ^= D[2];
+    A[3][2] ^= D[2];
+    A[4][2] ^= D[2];
+
+    A[0][3] ^= C[2];
+    A[1][3] ^= C[2];
+    A[2][3] ^= C[2];
+    A[3][3] ^= C[2];
+    A[4][3] ^= C[2];
+
+    A[0][4] ^= C[3];
+    A[1][4] ^= C[3];
+    A[2][4] ^= C[3];
+    A[3][4] ^= C[3];
+    A[4][4] ^= C[3];
+
+    A[0][0] ^= C[4];
+    A[1][0] ^= C[4];
+    A[2][0] ^= C[4];
+    A[3][0] ^= C[4];
+    A[4][0] ^= C[4];
+
+    C[1] = A[0][1];
+    C[2] = A[0][2];
+    C[3] = A[0][3];
+    C[4] = A[0][4];
+
+    A[0][1] = ROL64(A[1][1], rhotates[1][1]);
+    A[0][2] = ROL64(A[2][2], rhotates[2][2]);
+    A[0][3] = ROL64(A[3][3], rhotates[3][3]);
+    A[0][4] = ROL64(A[4][4], rhotates[4][4]);
+
+    A[1][1] = ROL64(A[1][4], rhotates[1][4]);
+    A[2][2] = ROL64(A[2][3], rhotates[2][3]);
+    A[3][3] = ROL64(A[3][2], rhotates[3][2]);
+    A[4][4] = ROL64(A[4][1], rhotates[4][1]);
+
+    A[1][4] = ROL64(A[4][2], rhotates[4][2]);
+    A[2][3] = ROL64(A[3][4], rhotates[3][4]);
+    A[3][2] = ROL64(A[2][1], rhotates[2][1]);
+    A[4][1] = ROL64(A[1][3], rhotates[1][3]);
+
+    A[4][2] = ROL64(A[2][4], rhotates[2][4]);
+    A[3][4] = ROL64(A[4][3], rhotates[4][3]);
+    A[2][1] = ROL64(A[1][2], rhotates[1][2]);
+    A[1][3] = ROL64(A[3][1], rhotates[3][1]);
+
+    A[2][4] = ROL64(A[4][0], rhotates[4][0]);
+    A[4][3] = ROL64(A[3][0], rhotates[3][0]);
+    A[1][2] = ROL64(A[2][0], rhotates[2][0]);
+    A[3][1] = ROL64(A[1][0], rhotates[1][0]);
+
+    A[1][0] = ROL64(C[3],    rhotates[0][3]);
+    A[2][0] = ROL64(C[1],    rhotates[0][1]);
+    A[3][0] = ROL64(C[4],    rhotates[0][4]);
+    A[4][0] = ROL64(C[2],    rhotates[0][2]);
+
+    C[0] = A[0][0];
+    C[1] = A[1][0];
+    D[0] = A[0][1];
+    D[1] = A[1][1];
+
+    A[0][0] ^= (~A[0][1] & A[0][2]);
+    A[1][0] ^= (~A[1][1] & A[1][2]);
+    A[0][1] ^= (~A[0][2] & A[0][3]);
+    A[1][1] ^= (~A[1][2] & A[1][3]);
+    A[0][2] ^= (~A[0][3] & A[0][4]);
+    A[1][2] ^= (~A[1][3] & A[1][4]);
+    A[0][3] ^= (~A[0][4] & C[0]);
+    A[1][3] ^= (~A[1][4] & C[1]);
+    A[0][4] ^= (~C[0]    & D[0]);
+    A[1][4] ^= (~C[1]    & D[1]);
+
+    C[2] = A[2][0];
+    C[3] = A[3][0];
+    D[2] = A[2][1];
+    D[3] = A[3][1];
+
+    A[2][0] ^= (~A[2][1] & A[2][2]);
+    A[3][0] ^= (~A[3][1] & A[3][2]);
+    A[2][1] ^= (~A[2][2] & A[2][3]);
+    A[3][1] ^= (~A[3][2] & A[3][3]);
+    A[2][2] ^= (~A[2][3] & A[2][4]);
+    A[3][2] ^= (~A[3][3] & A[3][4]);
+    A[2][3] ^= (~A[2][4] & C[2]);
+    A[3][3] ^= (~A[3][4] & C[3]);
+    A[2][4] ^= (~C[2]    & D[2]);
+    A[3][4] ^= (~C[3]    & D[3]);
+
+    C[4] = A[4][0];
+    D[4] = A[4][1];
+
+    A[4][0] ^= (~A[4][1] & A[4][2]);
+    A[4][1] ^= (~A[4][2] & A[4][3]);
+    A[4][2] ^= (~A[4][3] & A[4][4]);
+    A[4][3] ^= (~A[4][4] & C[4]);
+    A[4][4] ^= (~C[4]    & D[4]);
+    A[0][0] ^= iotas[i];
+}
+
+void KeccakF1600(uint64_t A[5][5])
+{
+    size_t i;
+
+    for (i = 0; i < 24; i++) {
+        Round(A, i);
+    }
+}
+
 #elif defined(KECCAK_2X)
 /*
  * This implementation is variant of KECCAK_1X above with outer-most
@@ -815,10 +980,10 @@ static uint64_t BitDeinterleave(uint64_t Ai)
 /*
  * SHA3_absorb can be called multiple times, but at each invocation
  * largest multiple of |r| out of |len| bytes are processed. Then
- * remaining amount of bytes are returned. This is done to spare caller
- * trouble of calculating the largest multiple of |r|, effectively the
- * blocksize. It is commonly (1600 - 256*n)/8, e.g. 168, 136, 104, 72,
- * but can also be (1600 - 448)/8 = 144. All this means that message
+ * remaining amount of bytes is returned. This is done to spare caller
+ * trouble of calculating the largest multiple of |r|. |r| can be viewed
+ * as blocksize. It is commonly (1600 - 256*n)/8, e.g. 168, 136, 104,
+ * 72, but can also be (1600 - 448)/8 = 144. All this means that message
  * padding and intermediate sub-block buffering, byte- or bitwise, is
  * caller's reponsibility.
  */
@@ -903,6 +1068,11 @@ void SHA3_squeeze(uint64_t A[5][5], unsigned char *out, size_t len, size_t r)
         }
     }
 }
+#else
+size_t SHA3_absorb(uint64_t A[5][5], const unsigned char *inp, size_t len,
+                   size_t r);
+void SHA3_squeeze(uint64_t A[5][5], unsigned char *out, size_t len, size_t r);
+#endif
 
 #ifdef SELFTEST
 /*


More information about the openssl-commits mailing list