From ab9a57280d7d56a76a9e9a2d860fe205eb45221f Mon Sep 17 00:00:00 2001
From: Jun He <jun.he@arm.com>
Date: Fri, 9 Jul 2021 16:49:06 +0800
Subject: Fix SSE3 and BMI2 compile error

After SHUFFLE code blocks are refactored, "tmmintrin.h"
is missed, and bmi2 code part will have build failure
as type conflicts.

Signed-off-by: Jun He <jun.he@arm.com>
Change-Id: I7800cd7e050f4d349e5a227206b14b9c566e547f
---
 snappy-internal.h | 35 +++++++++++++++++++++++++++++++++--
 snappy.cc         | 21 ---------------------
 2 files changed, 33 insertions(+), 23 deletions(-)

diff --git a/snappy-internal.h b/snappy-internal.h
index f1aafa9..71d19c0 100644
--- a/snappy-internal.h
+++ b/snappy-internal.h
@@ -33,13 +33,44 @@
 
 #include "snappy-stubs-internal.h"
 
+#if !defined(SNAPPY_HAVE_BMI2)
+// __BMI2__ is defined by GCC and Clang. Visual Studio doesn't target BMI2
+// specifically, but it does define __AVX2__ when AVX2 support is available.
+// Fortunately, AVX2 was introduced in Haswell, just like BMI2.
+//
+// BMI2 is not defined as a subset of AVX2 (unlike SSSE3 and AVX above). So,
+// GCC and Clang can build code with AVX2 enabled but BMI2 disabled, in which
+// case issuing BMI2 instructions results in a compiler error.
+#if defined(__BMI2__) || (defined(_MSC_VER) && defined(__AVX2__))
+#define SNAPPY_HAVE_BMI2 1
+#else
+#define SNAPPY_HAVE_BMI2 0
+#endif
+#endif  // !defined(SNAPPY_HAVE_BMI2)
+
+#if SNAPPY_HAVE_BMI2
+// Please do not replace with <x86intrin.h>. or with headers that assume more
+// advanced SSE versions without checking with all the OWNERS.
+#include <immintrin.h>
+#endif
+
+#if SNAPPY_HAVE_SSSE3
+#include <tmmintrin.h>
+#endif
+#if SNAPPY_HAVE_NEON
+#include <arm_neon.h>
+#endif
+
 namespace snappy {
 namespace internal {
 
+#if (SNAPPY_HAVE_SSSE3 || SNAPPY_HAVE_NEON)
+#define SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE 1
+#endif
 #if SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE
 #if SNAPPY_HAVE_SSSE3
 using V128 = __m128i;
-#else
+#elif SNAPPY_HAVE_NEON
 using V128 = uint8x16_t;
 #endif
 
@@ -72,7 +103,7 @@ inline V128 V128_Shuffle(V128 input, V128 shuffle_mask) {
 
 inline V128 V128_DupChar(char c) { return _mm_set1_epi8(c); }
 
-#else
+#elif SNAPPY_HAVE_NEON
 inline V128 V128_Load(const V128* src) {
   return vld1q_u8(reinterpret_cast<const uint8_t*>(src));
 }
diff --git a/snappy.cc b/snappy.cc
index 3f446c6..3b0de12 100644
--- a/snappy.cc
+++ b/snappy.cc
@@ -30,27 +30,6 @@
 #include "snappy-sinksource.h"
 #include "snappy.h"
 
-#if !defined(SNAPPY_HAVE_BMI2)
-// __BMI2__ is defined by GCC and Clang. Visual Studio doesn't target BMI2
-// specifically, but it does define __AVX2__ when AVX2 support is available.
-// Fortunately, AVX2 was introduced in Haswell, just like BMI2.
-//
-// BMI2 is not defined as a subset of AVX2 (unlike SSSE3 and AVX above). So,
-// GCC and Clang can build code with AVX2 enabled but BMI2 disabled, in which
-// case issuing BMI2 instructions results in a compiler error.
-#if defined(__BMI2__) || (defined(_MSC_VER) && defined(__AVX2__))
-#define SNAPPY_HAVE_BMI2 1
-#else
-#define SNAPPY_HAVE_BMI2 0
-#endif
-#endif  // !defined(SNAPPY_HAVE_BMI2)
-
-#if SNAPPY_HAVE_BMI2
-// Please do not replace with <x86intrin.h>. or with headers that assume more
-// advanced SSE versions without checking with all the OWNERS.
-#include <immintrin.h>
-#endif
-
 #include <algorithm>
 #include <array>
 #include <cstddef>
-- 
cgit v1.2.1


From 734b32bfe3d85675652964d09b59eff33a00f557 Mon Sep 17 00:00:00 2001
From: Jun He <jun.he@arm.com>
Date: Fri, 9 Jul 2021 16:59:45 +0800
Subject: Add config and header file for NEON support

Signed-off-by: Jun He <jun.he@arm.com>
Change-Id: I3fade568ff92b4303387705f843d0051d5e88349
---
 CMakeLists.txt    | 9 +++++++++
 cmake/config.h.in | 3 +++
 2 files changed, 12 insertions(+)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 672561e..6eef485 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -178,6 +178,15 @@ int main() {
   return _bzhi_u32(0, 1);
 }" SNAPPY_HAVE_BMI2)
 
+check_cxx_source_compiles("
+#include <arm_neon.h>
+int main() {
+  uint8_t val = 3, dup[8];
+  uint8x16_t v = vld1q_dup_u8(&val);
+  vst1q_u8(dup, v);
+  return 0;
+}" SNAPPY_HAVE_NEON)
+
 include(CheckSymbolExists)
 check_symbol_exists("mmap" "sys/mman.h" HAVE_FUNC_MMAP)
 check_symbol_exists("sysconf" "unistd.h" HAVE_FUNC_SYSCONF)
diff --git a/cmake/config.h.in b/cmake/config.h.in
index 872bd3c..568f69c 100644
--- a/cmake/config.h.in
+++ b/cmake/config.h.in
@@ -49,6 +49,9 @@
 /* Define to 1 if you target processors with BMI2+ and have <bmi2intrin.h>. */
 #cmakedefine01 SNAPPY_HAVE_BMI2
 
+/* Define to 1 if you target processors with NEON and have <arm_neon.h>. */
+#cmakedefine01 SNAPPY_HAVE_NEON
+
 /* Define to 1 if your processor stores words with the most significant byte
    first (like Motorola and SPARC, unlike Intel and VAX). */
 #cmakedefine SNAPPY_IS_BIG_ENDIAN 1
-- 
cgit v1.2.1