Use vec_sld in VecAdd64 and VecSub64

Using the proper mask and a shift saves a load. Thanks to Steven Munroe for the guidance.

1 files changed, 23 insertions, 28 deletions

diff --git a/ppc_simd.h b/ppc_simd.h
index 232f9f21..91e4b2aa 100644
--- a/ppc_simd.h
+++ b/ppc_simd.h
@@ -1969,24 +1969,24 @@ inline uint32x4_p VecAdd64(const uint32x4_p& vec1, const uint32x4_p& vec2)
 #if defined(_ARCH_PWR8) && !defined(CRYPTOPP_DEBUG)

     return (uint32x4_p)vec_add((uint64x2_p)vec1, (uint64x2_p)vec2);

 #else

-    // The carry mask selects carries for elements 1 and 3 and sets

-    // remaining elements to 0. The mask also shifts the carried values

-    // left by 4 bytes so the carries are added to elements 0 and 2.

-

-    // Small optimization... We can avoid a zero vector {0,0,0,0} and the

-    // load by using an element that will always be 0. Bytes 1,2, 5,6, 9,10,

-    // 13,14 are zero because we are using a vector unsigned int. There are

-    // no carries into those bytes using a vector unsigned int. The

-    // code below uses byte 2 for the 0 value.

+    // The carry mask selects carrys for elements 1 and 3 and sets

+    // remaining elements to 0. The results is then shifted so the

+    // carried values are subtracted from elements 0 and 2.

 #if defined(CRYPTOPP_BIG_ENDIAN)

-    const uint8x16_p cmask = {4,5,6,7, 2,2,2,2, 12,13,14,15, 2,2,2,2};

+    const uint32x4_p zero = {0, 0, 0, 0};

+    const uint32x4_p mask = {0, 1, 0, 1};

 #else

-    const uint8x16_p cmask = {2,2,2,2, 0,1,2,3, 2,2,2,2, 8,9,10,11};

+    const uint32x4_p zero = {0, 0, 0, 0};

+    const uint32x4_p mask = {1, 0, 1, 0};

 #endif

 

+    // subc sets the compliment of borrow, so we have to un-compliment it

+    // using andc.

     uint32x4_p cy = vec_addc(vec1, vec2);

-    cy = vec_perm(cy, cy, cmask);

-    return vec_add(vec_add(vec1, vec2), cy);

+    uint32x4_p res = vec_add(vec1, vec2);

+    cy = vec_and(mask, cy);

+    cy = vec_sld (cy, zero, 4);

+    return vec_add(res, cy);

 #endif

 }

 

@@ -2035,28 +2035,23 @@ inline uint32x4_p VecSub64(const uint32x4_p& vec1, const uint32x4_p& vec2)
     return (uint32x4_p)vec_sub((uint64x2_p)vec1, (uint64x2_p)vec2);

 #else

     // The borrow mask selects borrows for elements 1 and 3 and sets

-    // remaining elements to 0. The mask also shifts the borrowed values

-    // left by 4 bytes so the borrows are subtracted from elements 0 and 2.

-

-    // Small optimization... We can avoid a zero vector {0,0,0,0} and the

-    // load by using an element that will always be 0. Bytes 1,2, 5,6, 9,10,

-    // 13,14 are zero because we are using a vector unsigned int. There are

-    // no borrows from those bytes using a vector unsigned int. The

-    // code below uses byte 2 for the 0 value.

+    // remaining elements to 0. The results is then shifted so the

+    // borrowed values are subtracted from elements 0 and 2.

 #if defined(CRYPTOPP_BIG_ENDIAN)

-    const uint8x16_p bmask = {4,5,6,7, 2,2,2,2, 12,13,14,15, 2,2,2,2};

-    const uint32x4_p amask = {1, 1, 1, 1};

+    const uint32x4_p zero = {0, 0, 0, 0};

+    const uint32x4_p mask = {0, 1, 0, 1};

 #else

-    const uint8x16_p bmask = {2,2,2,2, 0,1,2,3, 2,2,2,2, 8,9,10,11};

-    const uint32x4_p amask = {1, 1, 1, 1};

+    const uint32x4_p zero = {0, 0, 0, 0};

+    const uint32x4_p mask = {1, 0, 1, 0};

 #endif

 

     // subc sets the compliment of borrow, so we have to un-compliment it

     // using andc.

     uint32x4_p bw = vec_subc(vec1, vec2);

-    bw = vec_andc(amask, bw);

-    bw = vec_perm(bw, bw, bmask);

-    return vec_sub(vec_sub(vec1, vec2), bw);

+    uint32x4_p res = vec_sub(vec1, vec2);

+    bw = vec_andc(mask, bw);

+    bw = vec_sld (bw, zero, 4);

+    return vec_sub(res, bw);

 #endif

 }