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.

2 files changed, 23 insertions, 23 deletions

diff --git a/src/lib/block/serpent/serpent.cpp b/src/lib/block/serpent/serpent.cpp
index 93af81231..6e1d79766 100644
--- a/src/lib/block/serpent/serpent.cpp
+++ b/src/lib/block/serpent/serpent.cpp
@@ -22,11 +22,11 @@ namespace {
 */
 inline void transform(uint32_t& B0, uint32_t& B1, uint32_t& B2, uint32_t& B3)
    {
-   B0  = rotate_left(B0, 13);   B2  = rotate_left(B2, 3);
-   B1 ^= B0 ^ B2;               B3 ^= B2 ^ (B0 << 3);
-   B1  = rotate_left(B1, 1);    B3  = rotate_left(B3, 7);
-   B0 ^= B1 ^ B3;               B2 ^= B3 ^ (B1 << 7);
-   B0  = rotate_left(B0, 5);    B2  = rotate_left(B2, 22);
+   B0  = rotl<13>(B0);   B2  = rotl<3>(B2);
+   B1 ^= B0 ^ B2;        B3 ^= B2 ^ (B0 << 3);
+   B1  = rotl<1>(B1);    B3  = rotl<7>(B3);
+   B0 ^= B1 ^ B3;        B2 ^= B3 ^ (B1 << 7);
+   B0  = rotl<5>(B0);    B2  = rotl<22>(B2);
    }
 
 /*
@@ -34,11 +34,11 @@ inline void transform(uint32_t& B0, uint32_t& B1, uint32_t& B2, uint32_t& B3)
 */
 inline void i_transform(uint32_t& B0, uint32_t& B1, uint32_t& B2, uint32_t& B3)
    {
-   B2  = rotate_right(B2, 22);  B0  = rotate_right(B0, 5);
-   B2 ^= B3 ^ (B1 << 7);        B0 ^= B1 ^ B3;
-   B3  = rotate_right(B3, 7);   B1  = rotate_right(B1, 1);
-   B3 ^= B2 ^ (B0 << 3);        B1 ^= B0 ^ B2;
-   B2  = rotate_right(B2, 3);   B0  = rotate_right(B0, 13);
+   B2  = rotr<22>(B2);   B0  = rotr<5>(B0);
+   B2 ^= B3 ^ (B1 << 7); B0 ^= B1 ^ B3;
+   B3  = rotr<7>(B3);    B1  = rotr<1>(B1);
+   B3 ^= B2 ^ (B0 << 3); B1 ^= B0 ^ B2;
+   B2  = rotr<3>(B2);    B0  = rotr<13>(B0);
    }
 
 }
@@ -192,7 +192,7 @@ void Serpent::key_schedule(const uint8_t key[], size_t length)
    for(size_t i = 8; i != 140; ++i)
       {
       uint32_t wi = W[i-8] ^ W[i-5] ^ W[i-3] ^ W[i-1] ^ PHI ^ uint32_t(i-8);
-      W[i] = rotate_left(wi, 11);
+      W[i] = rotl<11>(wi);
       }
 
    SBoxE1(W[ 20],W[ 21],W[ 22],W[ 23]);
diff --git a/src/lib/block/serpent/serpent_simd/serpent_simd.cpp b/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
index 94b3cf9ad..b184b0d4a 100644
--- a/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
+++ b/src/lib/block/serpent/serpent_simd/serpent_simd.cpp
@@ -24,30 +24,30 @@ namespace Botan {
 */
 #define transform(B0, B1, B2, B3)                                  \
    do {                                                            \
-      B0.rotate_left(13);                                          \
-      B2.rotate_left(3);                                           \
+      B0 = B0.rotl<13>();                                          \
+      B2 = B2.rotl<3>();                                           \
       B1 ^= B0 ^ B2;                                               \
       B3 ^= B2 ^ (B0 << 3);                                        \
-      B1.rotate_left(1);                                           \
-      B3.rotate_left(7);                                           \
+      B1 = B1.rotl<1>();                                           \
+      B3 = B3.rotl<7>();                                           \
       B0 ^= B1 ^ B3;                                               \
       B2 ^= B3 ^ (B1 << 7);                                        \
-      B0.rotate_left(5);                                           \
-      B2.rotate_left(22);                                          \
+      B0 = B0.rotl<5>();                                           \
+      B2 = B2.rotl<22>();                                          \
    } while(0);
 
 #define i_transform(B0, B1, B2, B3)                                \
    do {                                                            \
-      B2.rotate_right(22);                                         \
-      B0.rotate_right(5);                                          \
+      B2 = B2.rotr<22>();                                          \
+      B0 = B0.rotr<5>();                                           \
       B2 ^= B3 ^ (B1 << 7);                                        \
       B0 ^= B1 ^ B3;                                               \
-      B3.rotate_right(7);                                          \
-      B1.rotate_right(1);                                          \
+      B3 = B3.rotr<7>();                                           \
+      B1 = B1.rotr<1>();                                           \
       B3 ^= B2 ^ (B0 << 3);                                        \
       B1 ^= B0 ^ B2;                                               \
-      B2.rotate_right(3);                                          \
-      B0.rotate_right(13);                                         \
+      B2 = B2.rotr<3>();                                           \
+      B0 = B0.rotr<13>();                                          \
    } while(0);
 
 /*