Remove unrolled loops

[openssl.git] / crypto / ec / curve448 / f_generic.c

diff --git a/crypto/ec/curve448/f_generic.c b/crypto/ec/curve448/f_generic.c
index 135bab4828bb588fce901644bef812b57d80e8ba..6babea6e41717776d822d71821993d61c077b0f0 100644 (file)
--- a/crypto/ec/curve448/f_generic.c
+++ b/crypto/ec/curve448/f_generic.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2017 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2017-2018 The OpenSSL Project Authors. All Rights Reserved.
  * Copyright 2015-2016 Cryptography Research, Inc.
  *
  * Licensed under the OpenSSL license (the "License").  You may not use
@@ -17,12 +17,12 @@ static const gf MODULUS = {
                   0xffffffffffffff, 0xffffffffffffff)
 };
 
-/** Serialize to wire format. */
+/* Serialize to wire format. */
 void gf_serialize(uint8_t serial[SER_BYTES], const gf x, int with_hibit)
 {
     unsigned int j = 0, fill = 0;
     dword_t buffer = 0;
-    unsigned int i;
+    int i;
     gf red;
 
     gf_copy(red, x);
@@ -30,13 +30,13 @@ void gf_serialize(uint8_t serial[SER_BYTES], const gf x, int with_hibit)
     if (!with_hibit)
         assert(gf_hibit(red) == 0);
 
-    UNROLL for (i = 0; i < (with_hibit ? X_SER_BYTES : SER_BYTES); i++) {
+    for (i = 0; i < (with_hibit ? X_SER_BYTES : SER_BYTES); i++) {
         if (fill < 8 && j < NLIMBS) {
             buffer |= ((dword_t) red->limb[LIMBPERM(j)]) << fill;
             fill += LIMB_PLACE_VALUE(LIMBPERM(j));
             j++;
         }
-        serial[i] = buffer;
+        serial[i] = (uint8_t)buffer;
         fill -= 8;
         buffer >>= 8;
     }
@@ -46,18 +46,20 @@ void gf_serialize(uint8_t serial[SER_BYTES], const gf x, int with_hibit)
 mask_t gf_hibit(const gf x)
 {
     gf y;
+
     gf_add(y, x, x);
     gf_strong_reduce(y);
-    return -(y->limb[0] & 1);
+    return 0 - (y->limb[0] & 1);
 }
 
 /* Return high bit of x = low bit of 2x mod p */
 mask_t gf_lobit(const gf x)
 {
     gf y;
+
     gf_copy(y, x);
     gf_strong_reduce(y);
-    return -(y->limb[0] & 1);
+    return 0 - (y->limb[0] & 1);
 }
 
 /* Deserialize from wire format; return -1 on success and 0 on failure. */
@@ -71,25 +73,27 @@ mask_t gf_deserialize(gf x, const uint8_t serial[SER_BYTES], int with_hibit,
     unsigned int i;
     mask_t succ;
 
-    UNROLL for (i = 0; i < NLIMBS; i++) {
-        UNROLL while (fill < LIMB_PLACE_VALUE(LIMBPERM(i)) && j < nbytes) {
-            uint8_t sj = serial[j];
+    for (i = 0; i < NLIMBS; i++) {
+        while (fill < LIMB_PLACE_VALUE(LIMBPERM(i)) && j < nbytes) {
+            uint8_t sj;
+
+            sj = serial[j];
             if (j == nbytes - 1)
                 sj &= ~hi_nmask;
             buffer |= ((dword_t) sj) << fill;
             fill += 8;
             j++;
         }
-        x->limb[LIMBPERM(i)] =
-            (i < NLIMBS - 1) ? buffer & LIMB_MASK(LIMBPERM(i)) : buffer;
+        x->limb[LIMBPERM(i)] = (word_t)
+            ((i < NLIMBS - 1) ? buffer & LIMB_MASK(LIMBPERM(i)) : buffer);
         fill -= LIMB_PLACE_VALUE(LIMBPERM(i));
         buffer >>= LIMB_PLACE_VALUE(LIMBPERM(i));
         scarry =
             (scarry + x->limb[LIMBPERM(i)] -
              MODULUS->limb[LIMBPERM(i)]) >> (8 * sizeof(word_t));
     }
-    succ = with_hibit ? -(mask_t) 1 : ~gf_hibit(x);
-    return succ & word_is_zero(buffer) & ~word_is_zero(scarry);
+    succ = with_hibit ? 0 - (mask_t) 1 : ~gf_hibit(x);
+    return succ & word_is_zero((word_t)buffer) & ~word_is_zero((word_t)scarry);
 }
 
 /* Reduce to canonical form. */
@@ -120,7 +124,7 @@ void gf_strong_reduce(gf a)
      */
     assert(word_is_zero(scarry) | word_is_zero(scarry + 1));
 
-    scarry_0 = scarry;
+    scarry_0 = (word_t)scarry;
 
     /* add it back */
     for (i = 0; i < NLIMBS; i++) {
@@ -159,9 +163,42 @@ mask_t gf_eq(const gf a, const gf b)
     gf_sub(c, a, b);
     gf_strong_reduce(c);
 
-    for (i = 0; i < NLIMBS; i++) {
+    for (i = 0; i < NLIMBS; i++)
         ret |= c->limb[LIMBPERM(i)];
-    }
 
     return word_is_zero(ret);
 }
+
+mask_t gf_isr(gf a, const gf x)
+{
+    gf L0, L1, L2;
+
+    gf_sqr(L1, x);
+    gf_mul(L2, x, L1);
+    gf_sqr(L1, L2);
+    gf_mul(L2, x, L1);
+    gf_sqrn(L1, L2, 3);
+    gf_mul(L0, L2, L1);
+    gf_sqrn(L1, L0, 3);
+    gf_mul(L0, L2, L1);
+    gf_sqrn(L2, L0, 9);
+    gf_mul(L1, L0, L2);
+    gf_sqr(L0, L1);
+    gf_mul(L2, x, L0);
+    gf_sqrn(L0, L2, 18);
+    gf_mul(L2, L1, L0);
+    gf_sqrn(L0, L2, 37);
+    gf_mul(L1, L2, L0);
+    gf_sqrn(L0, L1, 37);
+    gf_mul(L1, L2, L0);
+    gf_sqrn(L0, L1, 111);
+    gf_mul(L2, L1, L0);
+    gf_sqr(L0, L2);
+    gf_mul(L1, x, L0);
+    gf_sqrn(L0, L1, 223);
+    gf_mul(L1, L2, L0);
+    gf_sqr(L2, L1);
+    gf_mul(L0, L2, x);
+    gf_copy(a, L1);
+    return gf_eq(L0, ONE);
+}