Revert changes for lgtm findings

This broke SunCC to the point of no repair. SunCC is using AVX2 instructions for C++ and SSE2. Man this compiler sucks...

8 files changed, 118 insertions, 227 deletions

diff --git a/Filelist.txt b/Filelist.txt
index 141ec0e8..933d1226 100644
--- a/Filelist.txt
+++ b/Filelist.txt
@@ -358,7 +358,6 @@ square.cpp
 square.h

 squaretb.cpp

 sse_simd.cpp

-sse_simd.h

 stdcpp.h

 strciphr.cpp

 strciphr.h

diff --git a/asn.cpp b/asn.cpp
index 1631f52e..ec5a097d 100644
--- a/asn.cpp
+++ b/asn.cpp
@@ -395,25 +395,25 @@ void EncodedObjectFilter::Put(const byte *inString, size_t length)
 }

 

 BERGeneralDecoder::BERGeneralDecoder(BufferedTransformation &inQueue)

-	: m_inQueue(inQueue), m_length(0), m_finished(false)

+	: m_inQueue(inQueue), m_finished(false)

 {

 	Init(DefaultTag);

 }

 

 BERGeneralDecoder::BERGeneralDecoder(BufferedTransformation &inQueue, byte asnTag)

-	: m_inQueue(inQueue), m_length(0), m_finished(false)

+	: m_inQueue(inQueue), m_finished(false)

 {

 	Init(asnTag);

 }

 

 BERGeneralDecoder::BERGeneralDecoder(BERGeneralDecoder &inQueue)

-	: m_inQueue(inQueue), m_length(0), m_finished(false)

+	: m_inQueue(inQueue), m_finished(false)

 {

 	Init(DefaultTag);

 }

 

 BERGeneralDecoder::BERGeneralDecoder(BERGeneralDecoder &inQueue, byte asnTag)

-	: m_inQueue(inQueue), m_length(0), m_finished(false)

+	: m_inQueue(inQueue), m_finished(false)

 {

 	Init(asnTag);

 }

@@ -514,22 +514,22 @@ lword BERGeneralDecoder::ReduceLength(lword delta)
 }

 

 DERGeneralEncoder::DERGeneralEncoder(BufferedTransformation &outQueue)

-	: m_outQueue(outQueue), m_asnTag(DefaultTag), m_finished(false)

+	: ByteQueue(), m_outQueue(outQueue), m_asnTag(DefaultTag), m_finished(false)

 {

 }

 

 DERGeneralEncoder::DERGeneralEncoder(BufferedTransformation &outQueue, byte asnTag)

-	: m_outQueue(outQueue), m_asnTag(asnTag), m_finished(false)

+	: ByteQueue(), m_outQueue(outQueue), m_asnTag(asnTag), m_finished(false)

 {

 }

 

 DERGeneralEncoder::DERGeneralEncoder(DERGeneralEncoder &outQueue)

-	: m_outQueue(outQueue), m_asnTag(DefaultTag), m_finished(false)

+	: ByteQueue(), m_outQueue(outQueue), m_asnTag(DefaultTag), m_finished(false)

 {

 }

 

 DERGeneralEncoder::DERGeneralEncoder(DERGeneralEncoder &outQueue, byte asnTag)

-	: m_outQueue(outQueue), m_asnTag(asnTag), m_finished(false)

+	: ByteQueue(), m_outQueue(outQueue), m_asnTag(asnTag), m_finished(false)

 {

 }

 

diff --git a/chacha_avx.cpp b/chacha_avx.cpp
index cdf50266..20693488 100644
--- a/chacha_avx.cpp
+++ b/chacha_avx.cpp
@@ -24,12 +24,18 @@
 # include <xmmintrin.h>

 # include <emmintrin.h>

 # include <immintrin.h>

-# include "sse_simd.h"

 #endif

 

 // Squash MS LNK4221 and libtool warnings

 extern const char CHACHA_AVX_FNAME[] = __FILE__;

 

+// Sun Studio 12.4 OK, 12.5 and 12.6 compile error.

+#if (__SUNPRO_CC >= 0x5140) && (__SUNPRO_CC <= 0x5150)

+# define MAYBE_CONST

+#else

+# define MAYBE_CONST const

+#endif

+

 // VS2017 and global optimization bug. TODO, figure out when

 // we can re-enable full optimizations for VS2017. Also see

 // https://github.com/weidai11/cryptopp/issues/649 and

@@ -85,10 +91,14 @@ NAMESPACE_BEGIN(CryptoPP)
 

 void ChaCha_OperateKeystream_AVX2(const word32 *state, const byte* input, byte *output, unsigned int rounds)

 {

-    const __m256i state0 = _mm256_broadcastsi128_si256(load_m128i<0>(state));

-    const __m256i state1 = _mm256_broadcastsi128_si256(load_m128i<1>(state));

-    const __m256i state2 = _mm256_broadcastsi128_si256(load_m128i<2>(state));

-    const __m256i state3 = _mm256_broadcastsi128_si256(load_m128i<3>(state));

+    MAYBE_CONST __m128i* state_mm = (MAYBE_CONST __m128i*)(state);

+    MAYBE_CONST __m256i* input_mm = (MAYBE_CONST __m256i*)(input);

+    __m256i* output_mm = reinterpret_cast<__m256i*>(output);

+

+    const __m256i state0 = _mm256_broadcastsi128_si256(_mm_loadu_si128(state_mm + 0));

+    const __m256i state1 = _mm256_broadcastsi128_si256(_mm_loadu_si128(state_mm + 1));

+    const __m256i state2 = _mm256_broadcastsi128_si256(_mm_loadu_si128(state_mm + 2));

+    const __m256i state3 = _mm256_broadcastsi128_si256(_mm_loadu_si128(state_mm + 3));

 

     const __m256i CTR0 = _mm256_set_epi32(0, 0, 0, 0, 0, 0, 0, 4);

     const __m256i CTR1 = _mm256_set_epi32(0, 0, 0, 1, 0, 0, 0, 5);

@@ -294,80 +304,80 @@ void ChaCha_OperateKeystream_AVX2(const word32 *state, const byte* input, byte *
     X3_3 = _mm256_add_epi32(X3_3, state3);

     X3_3 = _mm256_add_epi64(X3_3, CTR3);

 

-    if (input)

+    if (input_mm)

     {

-        store_m256i<0>(output, _mm256_xor_si256(load_m256i<0>(input),

+        _mm256_storeu_si256(output_mm + 0, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 0),

             _mm256_permute2x128_si256(X0_0, X0_1, 1 + (3 << 4))));

-        store_m256i<1>(output, _mm256_xor_si256(load_m256i<1>(input),

+        _mm256_storeu_si256(output_mm + 1, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 1),

             _mm256_permute2x128_si256(X0_2, X0_3, 1 + (3 << 4))));

-        store_m256i<2>(output, _mm256_xor_si256(load_m256i<2>(input),

+        _mm256_storeu_si256(output_mm + 2, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 2),

             _mm256_permute2x128_si256(X1_0, X1_1, 1 + (3 << 4))));

-        store_m256i<3>(output, _mm256_xor_si256(load_m256i<3>(input),

+        _mm256_storeu_si256(output_mm + 3, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 3),

             _mm256_permute2x128_si256(X1_2, X1_3, 1 + (3 << 4))));

     }

     else

     {

-        store_m256i<0>(output, _mm256_permute2x128_si256(X0_0, X0_1, 1 + (3 << 4)));

-        store_m256i<1>(output, _mm256_permute2x128_si256(X0_2, X0_3, 1 + (3 << 4)));

-        store_m256i<2>(output, _mm256_permute2x128_si256(X1_0, X1_1, 1 + (3 << 4)));

-        store_m256i<3>(output, _mm256_permute2x128_si256(X1_2, X1_3, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 0, _mm256_permute2x128_si256(X0_0, X0_1, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 1, _mm256_permute2x128_si256(X0_2, X0_3, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 2, _mm256_permute2x128_si256(X1_0, X1_1, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 3, _mm256_permute2x128_si256(X1_2, X1_3, 1 + (3 << 4)));

     }

 

-    if (input)

+    if (input_mm)

     {

-        store_m256i<4>(output, _mm256_xor_si256(load_m256i<4>(input),

+        _mm256_storeu_si256(output_mm + 4, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 4),

             _mm256_permute2x128_si256(X2_0, X2_1, 1 + (3 << 4))));

-        store_m256i<5>(output, _mm256_xor_si256(load_m256i<5>(input),

+        _mm256_storeu_si256(output_mm + 5, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 5),

             _mm256_permute2x128_si256(X2_2, X2_3, 1 + (3 << 4))));

-        store_m256i<6>(output, _mm256_xor_si256(load_m256i<6>(input),

+        _mm256_storeu_si256(output_mm + 6, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 6),

             _mm256_permute2x128_si256(X3_0, X3_1, 1 + (3 << 4))));

-        store_m256i<7>(output, _mm256_xor_si256(load_m256i<7>(input),

+        _mm256_storeu_si256(output_mm + 7, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 7),

             _mm256_permute2x128_si256(X3_2, X3_3, 1 + (3 << 4))));

     }

     else

     {

-        store_m256i<4>(output, _mm256_permute2x128_si256(X2_0, X2_1, 1 + (3 << 4)));

-        store_m256i<5>(output, _mm256_permute2x128_si256(X2_2, X2_3, 1 + (3 << 4)));

-        store_m256i<6>(output, _mm256_permute2x128_si256(X3_0, X3_1, 1 + (3 << 4)));

-        store_m256i<7>(output, _mm256_permute2x128_si256(X3_2, X3_3, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 4, _mm256_permute2x128_si256(X2_0, X2_1, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 5, _mm256_permute2x128_si256(X2_2, X2_3, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 6, _mm256_permute2x128_si256(X3_0, X3_1, 1 + (3 << 4)));

+        _mm256_storeu_si256(output_mm + 7, _mm256_permute2x128_si256(X3_2, X3_3, 1 + (3 << 4)));

     }

 

-    if (input)

+    if (input_mm)

     {

-        store_m256i<8>(output, _mm256_xor_si256(load_m256i<8>(input),

+        _mm256_storeu_si256(output_mm +  8, _mm256_xor_si256(_mm256_loadu_si256(input_mm +  8),

             _mm256_permute2x128_si256(X0_0, X0_1, 0 + (2 << 4))));

-        store_m256i<9>(output, _mm256_xor_si256(load_m256i<9>(input),

+        _mm256_storeu_si256(output_mm +  9, _mm256_xor_si256(_mm256_loadu_si256(input_mm +  9),

             _mm256_permute2x128_si256(X0_2, X0_3, 0 + (2 << 4))));

-        store_m256i<10>(output, _mm256_xor_si256(load_m256i<10>(input),

+        _mm256_storeu_si256(output_mm + 10, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 10),

             _mm256_permute2x128_si256(X1_0, X1_1, 0 + (2 << 4))));

-        store_m256i<11>(output, _mm256_xor_si256(load_m256i<11>(input),

+        _mm256_storeu_si256(output_mm + 11, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 11),

             _mm256_permute2x128_si256(X1_2, X1_3, 0 + (2 << 4))));

     }

     else

     {

-        store_m256i<8>(output, _mm256_permute2x128_si256(X0_0, X0_1, 0 + (2 << 4)));

-        store_m256i<9>(output, _mm256_permute2x128_si256(X0_2, X0_3, 0 + (2 << 4)));

-        store_m256i<10>(output, _mm256_permute2x128_si256(X1_0, X1_1, 0 + (2 << 4)));

-        store_m256i<11>(output, _mm256_permute2x128_si256(X1_2, X1_3, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm +  8, _mm256_permute2x128_si256(X0_0, X0_1, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm +  9, _mm256_permute2x128_si256(X0_2, X0_3, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 10, _mm256_permute2x128_si256(X1_0, X1_1, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 11, _mm256_permute2x128_si256(X1_2, X1_3, 0 + (2 << 4)));

     }

 

-    if (input)

+    if (input_mm)

     {

-        store_m256i<12>(output, _mm256_xor_si256(load_m256i<12>(input),

+        _mm256_storeu_si256(output_mm + 12, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 12),

             _mm256_permute2x128_si256(X2_0, X2_1, 0 + (2 << 4))));

-        store_m256i<13>(output, _mm256_xor_si256(load_m256i<13>(input),

+        _mm256_storeu_si256(output_mm + 13, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 13),

             _mm256_permute2x128_si256(X2_2, X2_3, 0 + (2 << 4))));

-        store_m256i<14>(output, _mm256_xor_si256(load_m256i<14>(input),

+        _mm256_storeu_si256(output_mm + 14, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 14),

             _mm256_permute2x128_si256(X3_0, X3_1, 0 + (2 << 4))));

-        store_m256i<15>(output, _mm256_xor_si256(load_m256i<15>(input),

+        _mm256_storeu_si256(output_mm + 15, _mm256_xor_si256(_mm256_loadu_si256(input_mm + 15),

             _mm256_permute2x128_si256(X3_2, X3_3, 0 + (2 << 4))));

     }

     else

     {

-        store_m256i<12>(output, _mm256_permute2x128_si256(X2_0, X2_1, 0 + (2 << 4)));

-        store_m256i<13>(output, _mm256_permute2x128_si256(X2_2, X2_3, 0 + (2 << 4)));

-        store_m256i<14>(output, _mm256_permute2x128_si256(X3_0, X3_1, 0 + (2 << 4)));

-        store_m256i<15>(output, _mm256_permute2x128_si256(X3_2, X3_3, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 12, _mm256_permute2x128_si256(X2_0, X2_1, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 13, _mm256_permute2x128_si256(X2_2, X2_3, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 14, _mm256_permute2x128_si256(X3_0, X3_1, 0 + (2 << 4)));

+        _mm256_storeu_si256(output_mm + 15, _mm256_permute2x128_si256(X3_2, X3_3, 0 + (2 << 4)));

     }

 

     // https://software.intel.com/en-us/articles/avoiding-avx-sse-transition-penalties

diff --git a/chacha_simd.cpp b/chacha_simd.cpp
index e225579d..9fd6b0f1 100644
--- a/chacha_simd.cpp
+++ b/chacha_simd.cpp
@@ -38,7 +38,6 @@
 #if (CRYPTOPP_SSE2_INTRIN_AVAILABLE)

 # include <xmmintrin.h>

 # include <emmintrin.h>

-# include "sse_simd.h"

 #endif

 

 #if defined(__SSSE3__)

@@ -566,10 +565,14 @@ void ChaCha_OperateKeystream_NEON(const word32 *state, const byte* input, byte *
 

 void ChaCha_OperateKeystream_SSE2(const word32 *state, const byte* input, byte *output, unsigned int rounds)

 {

-    const __m128i state0 = load_m128i<0>(state);

-    const __m128i state1 = load_m128i<1>(state);

-    const __m128i state2 = load_m128i<2>(state);

-    const __m128i state3 = load_m128i<3>(state);

+    const __m128i* state_mm = reinterpret_cast<const __m128i*>(state);

+    const __m128i* input_mm = reinterpret_cast<const __m128i*>(input);

+    __m128i* output_mm = reinterpret_cast<__m128i*>(output);

+

+    const __m128i state0 = _mm_load_si128(state_mm + 0);

+    const __m128i state1 = _mm_load_si128(state_mm + 1);

+    const __m128i state2 = _mm_load_si128(state_mm + 2);

+    const __m128i state3 = _mm_load_si128(state_mm + 3);

 

     __m128i r0_0 = state0;

     __m128i r0_1 = state1;

@@ -769,57 +772,57 @@ void ChaCha_OperateKeystream_SSE2(const word32 *state, const byte* input, byte *
     r3_3 = _mm_add_epi32(r3_3, state3);

     r3_3 = _mm_add_epi64(r3_3, _mm_set_epi32(0, 0, 0, 3));

 

-    if (input)

+    if (input_mm)

     {

-        r0_0 = _mm_xor_si128(load_m128i<0>(input), r0_0);

-        r0_1 = _mm_xor_si128(load_m128i<1>(input), r0_1);

-        r0_2 = _mm_xor_si128(load_m128i<2>(input), r0_2);

-        r0_3 = _mm_xor_si128(load_m128i<3>(input), r0_3);

+        r0_0 = _mm_xor_si128(_mm_loadu_si128(input_mm + 0), r0_0);

+        r0_1 = _mm_xor_si128(_mm_loadu_si128(input_mm + 1), r0_1);

+        r0_2 = _mm_xor_si128(_mm_loadu_si128(input_mm + 2), r0_2);

+        r0_3 = _mm_xor_si128(_mm_loadu_si128(input_mm + 3), r0_3);

     }

 

-    store_m128i<0>(output, r0_0);

-    store_m128i<1>(output, r0_1);

-    store_m128i<2>(output, r0_2);

-    store_m128i<3>(output, r0_3);

+    _mm_storeu_si128(output_mm + 0, r0_0);

+    _mm_storeu_si128(output_mm + 1, r0_1);

+    _mm_storeu_si128(output_mm + 2, r0_2);

+    _mm_storeu_si128(output_mm + 3, r0_3);

 

-    if (input)

+    if (input_mm)

     {

-        r1_0 = _mm_xor_si128(load_m128i<4>(input), r1_0);

-        r1_1 = _mm_xor_si128(load_m128i<5>(input), r1_1);

-        r1_2 = _mm_xor_si128(load_m128i<6>(input), r1_2);

-        r1_3 = _mm_xor_si128(load_m128i<7>(input), r1_3);

+        r1_0 = _mm_xor_si128(_mm_loadu_si128(input_mm + 4), r1_0);

+        r1_1 = _mm_xor_si128(_mm_loadu_si128(input_mm + 5), r1_1);

+        r1_2 = _mm_xor_si128(_mm_loadu_si128(input_mm + 6), r1_2);

+        r1_3 = _mm_xor_si128(_mm_loadu_si128(input_mm + 7), r1_3);

     }

 

-    store_m128i<4>(output, r1_0);

-    store_m128i<5>(output, r1_1);

-    store_m128i<6>(output, r1_2);

-    store_m128i<7>(output, r1_3);

+    _mm_storeu_si128(output_mm + 4, r1_0);

+    _mm_storeu_si128(output_mm + 5, r1_1);

+    _mm_storeu_si128(output_mm + 6, r1_2);

+    _mm_storeu_si128(output_mm + 7, r1_3);

 

-    if (input)

+    if (input_mm)

     {

-        r2_0 = _mm_xor_si128(load_m128i< 8>(input), r2_0);

-        r2_1 = _mm_xor_si128(load_m128i< 9>(input), r2_1);

-        r2_2 = _mm_xor_si128(load_m128i<10>(input), r2_2);

-        r2_3 = _mm_xor_si128(load_m128i<11>(input), r2_3);

+        r2_0 = _mm_xor_si128(_mm_loadu_si128(input_mm + 8), r2_0);

+        r2_1 = _mm_xor_si128(_mm_loadu_si128(input_mm + 9), r2_1);

+        r2_2 = _mm_xor_si128(_mm_loadu_si128(input_mm + 10), r2_2);

+        r2_3 = _mm_xor_si128(_mm_loadu_si128(input_mm + 11), r2_3);

     }

 

-    store_m128i< 8>(output, r2_0);

-    store_m128i< 9>(output, r2_1);

-    store_m128i<10>(output, r2_2);

-    store_m128i<11>(output, r2_3);

+    _mm_storeu_si128(output_mm + 8, r2_0);

+    _mm_storeu_si128(output_mm + 9, r2_1);

+    _mm_storeu_si128(output_mm + 10, r2_2);

+    _mm_storeu_si128(output_mm + 11, r2_3);

 

-    if (input)

+    if (input_mm)

     {

-        r3_0 = _mm_xor_si128(load_m128i<12>(input), r3_0);

-        r3_1 = _mm_xor_si128(load_m128i<13>(input), r3_1);

-        r3_2 = _mm_xor_si128(load_m128i<14>(input), r3_2);

-        r3_3 = _mm_xor_si128(load_m128i<15>(input), r3_3);

+        r3_0 = _mm_xor_si128(_mm_loadu_si128(input_mm + 12), r3_0);

+        r3_1 = _mm_xor_si128(_mm_loadu_si128(input_mm + 13), r3_1);

+        r3_2 = _mm_xor_si128(_mm_loadu_si128(input_mm + 14), r3_2);

+        r3_3 = _mm_xor_si128(_mm_loadu_si128(input_mm + 15), r3_3);

     }

 

-    store_m128i<12>(output, r3_0);

-    store_m128i<13>(output, r3_1);

-    store_m128i<14>(output, r3_2);

-    store_m128i<15>(output, r3_3);

+    _mm_storeu_si128(output_mm + 12, r3_0);

+    _mm_storeu_si128(output_mm + 13, r3_1);

+    _mm_storeu_si128(output_mm + 14, r3_2);

+    _mm_storeu_si128(output_mm + 15, r3_3);

 }

 

 #endif  // CRYPTOPP_SSE2_INTRIN_AVAILABLE

diff --git a/cryptlib.vcxproj b/cryptlib.vcxproj
index 1f81c046..4985c096 100644
--- a/cryptlib.vcxproj
+++ b/cryptlib.vcxproj
@@ -538,7 +538,6 @@
     <ClInclude Include="sosemanuk.h" />

     <ClInclude Include="speck.h" />

     <ClInclude Include="square.h" />

-    <ClInclude Include="sse_simd.h" />

     <ClInclude Include="stdcpp.h" />

     <ClInclude Include="strciphr.h" />

     <ClInclude Include="tea.h" />

diff --git a/cryptlib.vcxproj.filters b/cryptlib.vcxproj.filters
index 7b91f87b..de140e33 100644
--- a/cryptlib.vcxproj.filters
+++ b/cryptlib.vcxproj.filters
@@ -993,9 +993,6 @@
     <ClInclude Include="square.h">

       <Filter>Header Files</Filter>

     </ClInclude>

-    <ClInclude Include="sse_simd.h">

-      <Filter>Header Files</Filter>

-    </ClInclude>

     <ClInclude Include="stdcpp.h">

       <Filter>Header Files</Filter>

     </ClInclude>

diff --git a/gf2n_simd.cpp b/gf2n_simd.cpp
index 3bd3cd0e..1d4d933f 100644
--- a/gf2n_simd.cpp
+++ b/gf2n_simd.cpp
@@ -28,7 +28,6 @@
 #if (CRYPTOPP_CLMUL_AVAILABLE)

 # include <emmintrin.h>

 # include <wmmintrin.h>

-# include "sse_simd.h"

 #endif

 

 #if (CRYPTOPP_ARM_PMULL_AVAILABLE)

@@ -466,31 +465,36 @@ NAMESPACE_BEGIN(CryptoPP)
 void

 GF2NT_233_Multiply_Reduce_CLMUL(const word* pA, const word* pB, word* pC)

 {

-    __m128i a0 = load_m128i<0>(pA);

-    __m128i a1 = load_m128i<1>(pA);

-    __m128i b0 = load_m128i<0>(pB);

-    __m128i b1 = load_m128i<1>(pB);

+    const __m128i* pAA = reinterpret_cast<const __m128i*>(pA);

+    const __m128i* pBB = reinterpret_cast<const __m128i*>(pB);

+    __m128i a0 = _mm_loadu_si128(pAA+0);

+    __m128i a1 = _mm_loadu_si128(pAA+1);

+    __m128i b0 = _mm_loadu_si128(pBB+0);

+    __m128i b1 = _mm_loadu_si128(pBB+1);

 

     __m128i c0, c1, c2, c3;

     F2N_Multiply_256x256_CLMUL(c3, c2, c1, c0, a1, a0, b1, b0);

     GF2NT_233_Reduce_CLMUL(c3, c2, c1, c0);

 

-    store_m128i<0>(pC, c0);

-    store_m128i<1>(pC, c1);

+    __m128i* pCC = reinterpret_cast<__m128i*>(pC);

+    _mm_storeu_si128(pCC+0, c0);

+    _mm_storeu_si128(pCC+1, c1);

 }

 

 void

 GF2NT_233_Square_Reduce_CLMUL(const word* pA, word* pC)

 {

-    __m128i a0 = load_m128i<0>(pA);

-    __m128i a1 = load_m128i<1>(pA);

+    const __m128i* pAA = reinterpret_cast<const __m128i*>(pA);

+    __m128i a0 = _mm_loadu_si128(pAA+0);

+    __m128i a1 = _mm_loadu_si128(pAA+1);

 

     __m128i c0, c1, c2, c3;

     F2N_Square_256_CLMUL(c3, c2, c1, c0, a1, a0);

     GF2NT_233_Reduce_CLMUL(c3, c2, c1, c0);

 

-    store_m128i<0>(pC, c0);

-    store_m128i<1>(pC, c1);

+    __m128i* pCC = reinterpret_cast<__m128i*>(pC);

+    _mm_storeu_si128(pCC+0, c0);

+    _mm_storeu_si128(pCC+1, c1);

 }

 

 #elif (CRYPTOPP_ARM_PMULL_AVAILABLE)

diff --git a/sse_simd.h b/sse_simd.h
deleted file mode 100644
index fe3a0332..00000000
--- a/sse_simd.h
+++ /dev/null
@@ -1,121 +0,0 @@
-// sse_simd.h - written and placed in public domain by Jeffrey Walton

-//              Helper functions to work with SSE and above. The class file

-//              was added after a scan by lgtm.com. We caught some findings

-//              that were not problems, but we refactored to squash them.

-

-#ifndef CRYPTOPP_SSE_CRYPTO_H

-#define CRYPTOPP_SSE_CRYPTO_H

-

-#include "config.h"

-

-#if (CRYPTOPP_SSE2_INTRIN_AVAILABLE)

-# include <emmintrin.h>

-#endif

-

-#if (CRYPTOPP_AVX2_AVAILABLE)

-# include <immintrin.h>

-#endif

-

-NAMESPACE_BEGIN(CryptoPP)

-

-#if (CRYPTOPP_SSE2_INTRIN_AVAILABLE)

-

-template <unsigned int N>

-inline __m128i load_m128i(const byte* ptr)

-{

-    enum { SCALE=sizeof(__m128i)/sizeof(byte) };

-    return _mm_loadu_si128(

-        const_cast<__m128i*>(  // SunCC workaround

-        reinterpret_cast<const __m128i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m128i load_m128i(const word16* ptr)

-{

-    enum { SCALE=sizeof(__m128i)/sizeof(word16) };

-    return _mm_loadu_si128(

-        const_cast<__m128i*>(  // SunCC workaround

-        reinterpret_cast<const __m128i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m128i load_m128i(const word32* ptr)

-{

-    enum { SCALE=sizeof(__m128i)/sizeof(word32) };

-    return _mm_loadu_si128(

-        const_cast<__m128i*>(  // SunCC workaround

-        reinterpret_cast<const __m128i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m128i load_m128i(const word64* ptr)

-{

-    enum { SCALE=sizeof(__m128i)/sizeof(word64) };

-    return _mm_loadu_si128(

-        const_cast<__m128i*>(  // SunCC workaround

-        reinterpret_cast<const __m128i*>(ptr+SCALE*N)));

-}

-

-// N specifies the nth 128-bit element

-template <unsigned int N, class T>

-inline void store_m128i(T* ptr, __m128i val)

-{

-    enum { SCALE=sizeof(__m128i)/sizeof(T) };

-    return _mm_storeu_si128(

-        reinterpret_cast<__m128i*>(ptr+SCALE*N), val);

-}

-

-#endif

-

-#if (CRYPTOPP_AVX2_AVAILABLE)

-

-template <unsigned int N>

-inline __m256i load_m256i(const byte* ptr)

-{

-    enum { SCALE=sizeof(__m256i)/sizeof(byte) };

-    return _mm256_loadu_si256(

-        const_cast<__m256i*>(  // SunCC workaround

-        reinterpret_cast<const __m256i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m256i load_m256i(const word16* ptr)

-{

-    enum { SCALE=sizeof(__m256i)/sizeof(word16) };

-    return _mm256_loadu_si256(

-        const_cast<__m256i*>(  // SunCC workaround

-        reinterpret_cast<const __m256i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m256i load_m256i(const word32* ptr)

-{

-    enum { SCALE=sizeof(__m256i)/sizeof(word32) };

-    return _mm256_loadu_si256(

-        const_cast<__m256i*>(  // SunCC workaround

-        reinterpret_cast<const __m256i*>(ptr+SCALE*N)));

-}

-

-template <unsigned int N>

-inline __m256i load_m256i(const word64* ptr)

-{

-    enum { SCALE=sizeof(__m256i)/sizeof(word64) };

-    return _mm256_loadu_si256(

-        const_cast<__m256i*>(  // SunCC workaround

-        reinterpret_cast<const __m256i*>(ptr+SCALE*N)));

-}

-

-// N specifies the nth 256-bit element

-template <unsigned int N, class T>

-inline void store_m256i(T* ptr, __m256i val)

-{

-    enum { SCALE=sizeof(__m256i)/sizeof(T) };

-    return _mm256_storeu_si256(

-        reinterpret_cast<__m256i*>(ptr+SCALE*N), val);

-}

-

-#endif

-

-NAMESPACE_END

-

-#endif  // CRYPTOPP_SSE_CRYPTO_H