Add VecMergeHi and VecMergeLo PowerPC wrappers

1 files changed, 23 insertions, 21 deletions

diff --git a/gf2n_simd.cpp b/gf2n_simd.cpp
index b0296b5f..1630615c 100644
--- a/gf2n_simd.cpp
+++ b/gf2n_simd.cpp
@@ -317,6 +317,8 @@ using CryptoPP::VecXor;
 using CryptoPP::VecAnd;

 

 using CryptoPP::VecPermute;

+using CryptoPP::VecMergeHi;

+using CryptoPP::VecMergeLo;

 using CryptoPP::VecShiftLeft;

 using CryptoPP::VecShiftRight;

 using CryptoPP::VecRotateLeftOctet;

@@ -335,11 +337,11 @@ inline uint64x2_p VMULL_00LE(const uint64x2_p& a, const uint64x2_p& b)
 {

     const uint64x2_p z={0};

 #if defined(__ibmxl__) || (defined(_AIX) && defined(__xlC__))

-    return VMULL2LE(__vpmsumd (vec_mergeh(z, a), vec_mergeh(z, b)));

+    return VMULL2LE(__vpmsumd (VecMergeHi(z, a), VecMergeHi(z, b)));

 #elif defined(__clang__)

-    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (vec_mergeh(z, a), vec_mergeh(z, b)));

+    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (VecMergeHi(z, a), VecMergeHi(z, b)));

 #else

-    return VMULL2LE(__builtin_crypto_vpmsumd (vec_mergeh(z, a), vec_mergeh(z, b)));

+    return VMULL2LE(__builtin_crypto_vpmsumd (VecMergeHi(z, a), VecMergeHi(z, b)));

 #endif

 }

 

@@ -348,11 +350,11 @@ inline uint64x2_p VMULL_11LE(const uint64x2_p& a, const uint64x2_p& b)
 {

     const uint64x2_p z={0};

 #if defined(__ibmxl__) || (defined(_AIX) && defined(__xlC__))

-    return VMULL2LE(__vpmsumd (vec_mergel(z, a), b));

+    return VMULL2LE(__vpmsumd (VecMergeLo(z, a), b));

 #elif defined(__clang__)

-    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (vec_mergel(z, a), b));

+    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (VecMergeLo(z, a), b));

 #else

-    return VMULL2LE(__builtin_crypto_vpmsumd (vec_mergel(z, a), b));

+    return VMULL2LE(__builtin_crypto_vpmsumd (VecMergeLo(z, a), b));

 #endif

 }

 

@@ -365,16 +367,16 @@ F2N_Multiply_128x128_POWER8(uint64x2_p& c1, uint64x2_p& c0, const uint64x2_p& a,
 

     c0 = VMULL_00LE(a, b);

     c1 = VMULL_11LE(a, b);

-    t1 = vec_mergel(a, a);

+    t1 = VecMergeLo(a, a);

     t1 = VecXor(a, t1);

-    t2 = vec_mergel(b, b);

+    t2 = VecMergeLo(b, b);

     t2 = VecXor(b, t2);

     t1 = VMULL_00LE(t1, t2);

     t1 = VecXor(c0, t1);

     t1 = VecXor(c1, t1);

     t2 = t1;

-    t1 = vec_mergeh(z0, t1);

-    t2 = vec_mergel(t2, z0);

+    t1 = VecMergeHi(z0, t1);

+    t2 = VecMergeLo(t2, z0);

     c0 = VecXor(c0, t1);

     c1 = VecXor(c1, t2);

 }

@@ -388,7 +390,7 @@ inline uint64x2_p ShiftLeft128_POWER8(uint64x2_p x)
 

     x = VecShiftLeft<N>(x);

     u = VecShiftRight<64-N>(u);

-    v = vec_mergeh(z, u);

+    v = VecMergeHi(z, u);

     x = VecOr(x, v);

     return x;

 }

@@ -405,41 +407,41 @@ GF2NT_233_Reduce_POWER8(uint64x2_p& c3, uint64x2_p& c2, uint64x2_p& c1, uint64x2
     const uint64x2_p z0={0};

 

     b1 = c1; a1 = c1;

-    a0 = vec_mergeh(c1, z0);

+    a0 = VecMergeHi(c1, z0);

     a1 = VecShiftLeft<23>(a1);

     a1 = VecShiftRight<23>(a1);

     c1 = VecOr(a1, a0);

     b2 = VecShiftRight<64-23>(c2);

     c3 = ShiftLeft128_POWER8<23>(c3);

-    a0 = vec_mergel(b2, z0);

+    a0 = VecMergeLo(b2, z0);

     c3 = VecOr(c3, a0);

     b1 = VecShiftRight<64-23>(b1);

     c2 = ShiftLeft128_POWER8<23>(c2);

-    a0 = vec_mergel(b1, z0);

+    a0 = VecMergeLo(b1, z0);

     c2 = VecOr(c2, a0);

     b3 = c3;

     b2 = VecShiftRight<64-10>(c2);

     b3 = ShiftLeft128_POWER8<10>(b3);

-    a0 = vec_mergel(b2, z0);

+    a0 = VecMergeLo(b2, z0);

     b3 = VecOr(b3, a0);

-    a0 = vec_mergel(c3, z0);

+    a0 = VecMergeLo(c3, z0);

     b3 = VecXor(b3, a0);

     b1 = VecShiftRight<64-23>(b3);

     b3 = ShiftLeft128_POWER8<23>(b3);

-    b3 = vec_mergel(b3, z0);

+    b3 = VecMergeLo(b3, z0);

     b3 = VecOr(b3, b1);

     c2 = VecXor(c2, b3);

     b3 = c3;

     b2 = VecShiftRight<64-10>(c2);

     b3 = ShiftLeft128_POWER8<10>(b3);

-    b2 = vec_mergel(b2, z0);

+    b2 = VecMergeLo(b2, z0);

     b3 = VecOr(b3, b2);

     b2 = c2;

     b2 = ShiftLeft128_POWER8<10>(b2);

-    a0 = vec_mergeh(z0, b2);

+    a0 = VecMergeHi(z0, b2);

     c2 = VecXor(c2, a0);

-    a0 = vec_mergeh(z0, b3);

-    a1 = vec_mergel(b2, z0);

+    a0 = VecMergeHi(z0, b3);

+    a1 = VecMergeLo(b2, z0);

     a0 = VecOr(a0, a1);

     c3 = VecXor(c3, a0);

     c0 = VecXor(c0, c2);