From 9b61d4143da890d23328c22db8d9454235e1ded7 Mon Sep 17 00:00:00 2001
From: Jeffrey Walton <noloader@gmail.com>
Date: Tue, 5 Dec 2017 12:32:26 -0500
Subject: [PATCH] Add big- and little-endian rotates for Aarch32 and Aarch64

---
 simon-simd.cpp | 24 ++++++++++++++++++++++++
 speck-simd.cpp | 24 ++++++++++++++++++++++++
 2 files changed, 48 insertions(+)

diff --git a/simon-simd.cpp b/simon-simd.cpp
index 7af34ac3..b9c4e337 100644
--- a/simon-simd.cpp
+++ b/simon-simd.cpp
@@ -86,8 +86,14 @@ inline uint32x4_t RotateRight32(const uint32x4_t& val)
 template <>
 inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 3,0,1,2, 7,4,5,6, 11,8,9,10, 15,12,13,14 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u32_u8(
         vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
 }
@@ -96,8 +102,14 @@ inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
 template <>
 inline uint32x4_t RotateRight32<8>(const uint32x4_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 1,2,3,0, 5,6,7,4, 9,10,11,8, 13,14,14,12 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u32_u8(
         vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
 }
@@ -521,8 +533,14 @@ inline uint64x2_t RotateRight64(const uint64x2_t& val)
 template <>
 inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 7,0,1,2, 3,4,5,6, 15,8,9,10, 11,12,13,14 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u64_u8(
         vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
 }
@@ -531,8 +549,14 @@ inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
 template <>
 inline uint64x2_t RotateRight64<8>(const uint64x2_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 1,2,3,4, 5,6,7,0, 9,10,11,12, 13,14,15,8 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u64_u8(
         vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
 }
diff --git a/speck-simd.cpp b/speck-simd.cpp
index 41df95d4..e23d6aec 100644
--- a/speck-simd.cpp
+++ b/speck-simd.cpp
@@ -84,8 +84,14 @@ inline uint32x4_t RotateRight32(const uint32x4_t& val)
 template <>
 inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 14,13,12,15, 10,9,8,11, 6,5,4,7, 2,1,0,3 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 3,0,1,2, 7,4,5,6, 11,8,9,10, 15,12,13,14 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u32_u8(
         vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
 }
@@ -94,8 +100,14 @@ inline uint32x4_t RotateLeft32<8>(const uint32x4_t& val)
 template <>
 inline uint32x4_t RotateRight32<8>(const uint32x4_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 12,15,14,13, 8,11,10,9, 4,7,6,5, 0,3,2,1 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 1,2,3,0, 5,6,7,4, 9,10,11,8, 13,14,14,12 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u32_u8(
         vqtbl1q_u8(vreinterpretq_u8_u32(val), mask));
 }
@@ -491,8 +503,14 @@ inline uint64x2_t RotateRight64(const uint64x2_t& val)
 template <>
 inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 14,13,12,11, 10,9,8,15, 6,5,4,3, 2,1,0,7 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 7,0,1,2, 3,4,5,6, 15,8,9,10, 11,12,13,14 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u64_u8(
         vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
 }
@@ -501,8 +519,14 @@ inline uint64x2_t RotateLeft64<8>(const uint64x2_t& val)
 template <>
 inline uint64x2_t RotateRight64<8>(const uint64x2_t& val)
 {
+#if defined(CRYPTOPP_BIG_ENDIAN)
     const uint8_t maskb[16] = { 8,15,14,13, 12,11,10,9, 0,7,6,5, 4,3,2,1 };
     const uint8x16_t mask = vld1q_u8(maskb);
+#else
+    const uint8_t maskb[16] = { 1,2,3,4, 5,6,7,0, 9,10,11,12, 13,14,15,8 };
+    const uint8x16_t mask = vld1q_u8(maskb);
+#endif
+
     return vreinterpretq_u64_u8(
         vqtbl1q_u8(vreinterpretq_u8_u64(val), mask));
 }