Switch to vec_mergeh and vec_mergel for PowerPC in binary curves

This saves aboout 4 instructions in VMULL

1 files changed, 8 insertions, 8 deletions

diff --git a/gf2n_simd.cpp b/gf2n_simd.cpp
index 903d38c7..9fa9aec3 100644
--- a/gf2n_simd.cpp
+++ b/gf2n_simd.cpp
@@ -317,8 +317,6 @@ using CryptoPP::VecOr;
 using CryptoPP::VecXor;

 using CryptoPP::VecAnd;

 

-using CryptoPP::VecGetLow;

-using CryptoPP::VecGetHigh;

 using CryptoPP::VecPermute;

 using CryptoPP::VecShiftLeft;

 using CryptoPP::VecShiftRight;

@@ -336,24 +334,26 @@ inline uint64x2_p VMULL2LE(const uint64x2_p& val)
 // _mm_clmulepi64_si128(a, b, 0x00)

 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 (VecGetHigh(a), VecGetHigh(b)));

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

 #elif defined(__clang__)

-    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (VecGetHigh(a), VecGetHigh(b)));

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

 #else

-    return VMULL2LE(__builtin_crypto_vpmsumd (VecGetHigh(a), VecGetHigh(b)));

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

 #endif

 }

 

 // _mm_clmulepi64_si128(a, b, 0x11)

 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 (VecGetLow(a), b));

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

 #elif defined(__clang__)

-    return VMULL2LE(__builtin_altivec_crypto_vpmsumd (VecGetLow(a), b));

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

 #else

-    return VMULL2LE(__builtin_crypto_vpmsumd (VecGetLow(a), b));

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

 #endif

 }