sha/keccak1600.c: reduce temporary storage utilization even futher.

Reviewed-by: Rich Salz <rsalz@openssl.org>

1 files changed, 46 insertions, 47 deletions

diff --git a/crypto/sha/keccak1600.c b/crypto/sha/keccak1600.c
index 827affbad3..af24a5d2ed 100644
--- a/crypto/sha/keccak1600.c
+++ b/crypto/sha/keccak1600.c
@@ -343,9 +343,9 @@ void KeccakF1600(uint64_t A[5][5])
 
 #elif defined(KECCAK_1X_ALT)
 /*
- * This is variant of above KECCAK_1X that presses temporary storage
- * even further, but at cost of more write references to A[5][5].
- * It's less suitable if A[5][5] is memory bound, but better if it's
+ * 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.
  */
 
@@ -361,40 +361,41 @@ 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];
 
-    D[0]     = ROL64(C[1], 1) ^ C[4];
-    A[0][0] ^= D[0];
-    A[1][0] ^= D[0];
-    A[2][0] ^= D[0];
-    A[3][0] ^= D[0];
-    A[4][0] ^= D[0];
+    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);
 
-    D[1]     = ROL64(C[2], 1) ^ C[0];
     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];
 
-    D[2]     = ROL64(C[3], 1) ^ C[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];
 
-    D[3]     = ROL64(C[4], 1) ^ C[2];
-    A[0][3] ^= D[3];
-    A[1][3] ^= D[3];
-    A[2][3] ^= D[3];
-    A[3][3] ^= D[3];
-    A[4][3] ^= D[3];
+    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];
 
-    D[4]     = ROL64(C[0], 1) ^ C[3];
-    A[0][4] ^= D[4];
-    A[1][4] ^= D[4];
-    A[2][4] ^= D[4];
-    A[3][4] ^= D[4];
-    A[4][4] ^= D[4];
+    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];
@@ -433,46 +434,44 @@ static void Round(uint64_t A[5][5], size_t i)
 
     C[0] = A[0][0];
     C[1] = A[1][0];
-    C[2] = A[2][0];
-    C[3] = A[3][0];
-    C[4] = A[4][0];
-
     D[0] = A[0][1];
     D[1] = A[1][1];
-    D[2] = A[2][1];
-    D[3] = A[3][1];
-    D[4] = A[4][1];
 
     A[0][0] ^= (~A[0][1] & A[0][2]);
     A[1][0] ^= (~A[1][1] & A[1][2]);
-    A[2][0] ^= (~A[2][1] & A[2][2]);
-    A[3][0] ^= (~A[3][1] & A[3][2]);
-    A[4][0] ^= (~A[4][1] & A[4][2]);
-
     A[0][1] ^= (~A[0][2] & A[0][3]);
     A[1][1] ^= (~A[1][2] & A[1][3]);
-    A[2][1] ^= (~A[2][2] & A[2][3]);
-    A[3][1] ^= (~A[3][2] & A[3][3]);
-    A[4][1] ^= (~A[4][2] & A[4][3]);
-
     A[0][2] ^= (~A[0][3] & A[0][4]);
     A[1][2] ^= (~A[1][3] & A[1][4]);
-    A[2][2] ^= (~A[2][3] & A[2][4]);
-    A[3][2] ^= (~A[3][3] & A[3][4]);
-    A[4][2] ^= (~A[4][3] & A[4][4]);
-
     A[0][3] ^= (~A[0][4] & C[0]);
     A[1][3] ^= (~A[1][4] & C[1]);
-    A[2][3] ^= (~A[2][4] & C[2]);
-    A[3][3] ^= (~A[3][4] & C[3]);
-    A[4][3] ^= (~A[4][4] & C[4]);
-
     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]);
-    A[4][4] ^= (~C[4]    & D[4]);
 
+    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];
 }