Optimize and cleanup SHACAL2

The SHA-2 Ch and Maj functions can be done very fast on both
AltiVec and NEON.

Also, we can take advantage of the POWER8 SHA-2 extensions here.

2 files changed, 8 insertions, 8 deletions

diff --git a/src/lib/block/shacal2/shacal2_avx2/shacal2_avx2.cpp b/src/lib/block/shacal2/shacal2_avx2/shacal2_avx2.cpp
index cb83fece8..e9928d3c1 100644
--- a/src/lib/block/shacal2/shacal2_avx2/shacal2_avx2.cpp
+++ b/src/lib/block/shacal2/shacal2_avx2/shacal2_avx2.cpp
@@ -16,9 +16,9 @@ void BOTAN_FORCE_INLINE BOTAN_FUNC_ISA("avx2")
                const SIMD_8x32& E, const SIMD_8x32& F, const SIMD_8x32& G, SIMD_8x32& H,
                uint32_t RK)
    {
-   H += E.rho<6,11,25>() + ((E & F) ^ (~E & G)) + SIMD_8x32::splat(RK);
+   H += E.sigma1() + SIMD_8x32::choose(E, F, G) + SIMD_8x32::splat(RK);
    D += H;
-   H += A.rho<2,13,22>() + ((A & B) | ((A | B) & C));
+   H += A.sigma0() + SIMD_8x32::majority(A, B, C);
    }
 
 void BOTAN_FORCE_INLINE BOTAN_FUNC_ISA("avx2")
@@ -26,9 +26,9 @@ void BOTAN_FORCE_INLINE BOTAN_FUNC_ISA("avx2")
                const SIMD_8x32& E, const SIMD_8x32& F, const SIMD_8x32& G, SIMD_8x32& H,
                uint32_t RK)
    {
-   H -= A.rho<2,13,22>() + ((A & B) | ((A | B) & C));
+   H -= A.sigma0() + SIMD_8x32::majority(A, B, C);
    D -= H;
-   H -= E.rho<6,11,25>() + ((E & F) ^ (~E & G)) + SIMD_8x32::splat(RK);
+   H -= E.sigma1() + SIMD_8x32::choose(E, F, G) + SIMD_8x32::splat(RK);
    }
 
 }
diff --git a/src/lib/block/shacal2/shacal2_simd/shacal2_simd.cpp b/src/lib/block/shacal2/shacal2_simd/shacal2_simd.cpp
index c7b6182f2..7b541a1a4 100644
--- a/src/lib/block/shacal2/shacal2_simd/shacal2_simd.cpp
+++ b/src/lib/block/shacal2/shacal2_simd/shacal2_simd.cpp
@@ -17,9 +17,9 @@ void SHACAL2_Fwd(const SIMD_4x32& A, const SIMD_4x32& B, const SIMD_4x32& C, SIM
                  const SIMD_4x32& E, const SIMD_4x32& F, const SIMD_4x32& G, SIMD_4x32& H,
                  uint32_t RK)
    {
-   H += E.rho<6,11,25>() + ((E & F) ^ (~E & G)) + SIMD_4x32::splat(RK);
+   H += E.sigma1() + SIMD_4x32::choose(E, F, G) + SIMD_4x32::splat(RK);
    D += H;
-   H += A.rho<2,13,22>() + ((A & B) | ((A | B) & C));
+   H += A.sigma0() + SIMD_4x32::majority(A, B, C);
    }
 
 inline
@@ -27,9 +27,9 @@ void SHACAL2_Rev(const SIMD_4x32& A, const SIMD_4x32& B, const SIMD_4x32& C, SIM
                  const SIMD_4x32& E, const SIMD_4x32& F, const SIMD_4x32& G, SIMD_4x32& H,
                  uint32_t RK)
    {
-   H -= A.rho<2,13,22>() + ((A & B) | ((A | B) & C));
+   H -= A.sigma0() + SIMD_4x32::majority(A, B, C);
    D -= H;
-   H -= E.rho<6,11,25>() + ((E & F) ^ (~E & G)) + SIMD_4x32::splat(RK);
+   H -= E.sigma1() + SIMD_4x32::choose(E, F, G) + SIMD_4x32::splat(RK);
    }
 
 }