Add compile-time rotation functions

The problem with asm rol/ror is the compiler can't schedule effectively.
But we only need asm in the case when the rotation is variable, so distinguish
the two cases. If a compile time constant, then static_assert that the rotation
is in the correct range and do the straightforward expression knowing the compiler
will probably do the right thing. Otherwise do a tricky expression that both
GCC and Clang happen to have recognize. Avoid the reduction case; instead
require that the rotation be in range (this reverts 2b37c13dcf).

Remove the asm rotations (making this branch illnamed), because now both Clang
and GCC will create a roll without any extra help.

Remove the reduction/mask by the word size for the variable case. The compiler
can't optimize that it out well, but it's easy to ensure it is valid in the callers,
especially now that the variable input cases are easy to grep for.

1 files changed, 12 insertions, 12 deletions

diff --git a/src/lib/block/des/des.cpp b/src/lib/block/des/des.cpp
index 44f315047..15c2adb66 100644
--- a/src/lib/block/des/des.cpp
+++ b/src/lib/block/des/des.cpp
@@ -91,16 +91,16 @@ void des_encrypt(uint32_t& L, uint32_t& R,
       {
       uint32_t T0, T1;
 
-      T0 = rotate_right(R, 4) ^ round_key[2*i];
-      T1 =              R     ^ round_key[2*i + 1];
+      T0 = rotr<4>(R) ^ round_key[2*i];
+      T1 =         R  ^ round_key[2*i + 1];
 
       L ^= DES_SPBOX1[get_byte(0, T0)] ^ DES_SPBOX2[get_byte(0, T1)] ^
            DES_SPBOX3[get_byte(1, T0)] ^ DES_SPBOX4[get_byte(1, T1)] ^
            DES_SPBOX5[get_byte(2, T0)] ^ DES_SPBOX6[get_byte(2, T1)] ^
            DES_SPBOX7[get_byte(3, T0)] ^ DES_SPBOX8[get_byte(3, T1)];
 
-      T0 = rotate_right(L, 4) ^ round_key[2*i + 2];
-      T1 =              L     ^ round_key[2*i + 3];
+      T0 = rotr<4>(L) ^ round_key[2*i + 2];
+      T1 =         L  ^ round_key[2*i + 3];
 
       R ^= DES_SPBOX1[get_byte(0, T0)] ^ DES_SPBOX2[get_byte(0, T1)] ^
            DES_SPBOX3[get_byte(1, T0)] ^ DES_SPBOX4[get_byte(1, T1)] ^
@@ -119,16 +119,16 @@ void des_decrypt(uint32_t& L, uint32_t& R,
       {
       uint32_t T0, T1;
 
-      T0 = rotate_right(R, 4) ^ round_key[2*i - 2];
-      T1 =              R     ^ round_key[2*i - 1];
+      T0 = rotr<4>(R) ^ round_key[2*i - 2];
+      T1 =         R  ^ round_key[2*i - 1];
 
       L ^= DES_SPBOX1[get_byte(0, T0)] ^ DES_SPBOX2[get_byte(0, T1)] ^
            DES_SPBOX3[get_byte(1, T0)] ^ DES_SPBOX4[get_byte(1, T1)] ^
            DES_SPBOX5[get_byte(2, T0)] ^ DES_SPBOX6[get_byte(2, T1)] ^
            DES_SPBOX7[get_byte(3, T0)] ^ DES_SPBOX8[get_byte(3, T1)];
 
-      T0 = rotate_right(L, 4) ^ round_key[2*i - 4];
-      T1 =              L     ^ round_key[2*i - 3];
+      T0 = rotr<4>(L) ^ round_key[2*i - 4];
+      T1 =         L  ^ round_key[2*i - 3];
 
       R ^= DES_SPBOX1[get_byte(0, T0)] ^ DES_SPBOX2[get_byte(0, T1)] ^
            DES_SPBOX3[get_byte(1, T0)] ^ DES_SPBOX4[get_byte(1, T1)] ^
@@ -160,7 +160,7 @@ void DES::encrypt_n(const uint8_t in[], uint8_t out[], size_t blocks) const
           (DES_FPTAB1[get_byte(2, L)] << 1) | (DES_FPTAB2[get_byte(3, L)] << 1) |
           (DES_FPTAB1[get_byte(0, R)] << 4) | (DES_FPTAB1[get_byte(1, R)] << 2) |
           (DES_FPTAB1[get_byte(2, R)]     ) | (DES_FPTAB2[get_byte(3, R)]     );
-      T = rotate_left(T, 32);
+      T = rotl<32>(T);
 
       store_be(T, out + 8*i);
       }
@@ -188,7 +188,7 @@ void DES::decrypt_n(const uint8_t in[], uint8_t out[], size_t blocks) const
           (DES_FPTAB1[get_byte(0, R)] << 4) | (DES_FPTAB1[get_byte(1, R)] << 2) |
           (DES_FPTAB1[get_byte(2, R)]     ) | (DES_FPTAB2[get_byte(3, R)]     );
 
-      T = rotate_left(T, 32);
+      T = rotl<32>(T);
 
       store_be(T, out + BLOCK_SIZE*i);
       }
@@ -232,7 +232,7 @@ void TripleDES::encrypt_n(const uint8_t in[], uint8_t out[], size_t blocks) cons
           (DES_FPTAB1[get_byte(0, R)] << 4) | (DES_FPTAB1[get_byte(1, R)] << 2) |
           (DES_FPTAB1[get_byte(2, R)]     ) | (DES_FPTAB2[get_byte(3, R)]     );
 
-      T = rotate_left(T, 32);
+      T = rotl<32>(T);
 
       store_be(T, out);
 
@@ -265,7 +265,7 @@ void TripleDES::decrypt_n(const uint8_t in[], uint8_t out[], size_t blocks) cons
           (DES_FPTAB1[get_byte(0, R)] << 4) | (DES_FPTAB1[get_byte(1, R)] << 2) |
           (DES_FPTAB1[get_byte(2, R)]     ) | (DES_FPTAB2[get_byte(3, R)]     );
 
-      T = rotate_left(T, 32);
+      T = rotl<32>(T);
 
       store_be(T, out);