/*
- * Copyright 2015 Facebook, Inc.
+ * Copyright 2016 Facebook, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* limitations under the License.
*/
-#ifndef FOLLY_GROUPVARINT_H_
-#define FOLLY_GROUPVARINT_H_
+#pragma once
#if !defined(__GNUC__) && !defined(_MSC_VER)
#error GroupVarint.h requires GCC or MSVC
#include <folly/Portability.h>
-#if FOLLY_X64 || defined(__i386__)
+#if FOLLY_X64 || defined(__i386__) || FOLLY_PPC64 || FOLLY_A64
#define HAVE_GROUP_VARINT 1
#include <cstdint>
#include <folly/detail/GroupVarintDetail.h>
#include <folly/Bits.h>
#include <folly/Range.h>
+#include <folly/portability/Builtins.h>
#include <glog/logging.h>
-#ifdef __SSSE3__
-#include <x86intrin.h>
+#if FOLLY_SSE >= 3
+#include <nmmintrin.h>
namespace folly {
namespace detail {
-extern const __m128i groupVarintSSEMasks[];
+alignas(16) extern const uint64_t groupVarintSSEMasks[];
} // namespace detail
} // namespace folly
#endif
return decode_simple(p, dest, dest+1, dest+2, dest+3);
}
-#ifdef __SSSE3__
+#if FOLLY_SSE >= 3
/**
* Just like the non-SSSE3 decode below, but with the additional constraint
* that we must be able to read at least 17 bytes from the input pointer, p.
static const char* decode(const char* p, uint32_t* dest) {
uint8_t key = p[0];
__m128i val = _mm_loadu_si128((const __m128i*)(p+1));
- __m128i mask = detail::groupVarintSSEMasks[key];
+ __m128i mask =
+ _mm_load_si128((const __m128i*)&detail::groupVarintSSEMasks[key * 2]);
__m128i r = _mm_shuffle_epi8(val, mask);
_mm_storeu_si128((__m128i*)dest, r);
return p + detail::groupVarintLengths[key];
uint32_t* c, uint32_t* d) {
uint8_t key = p[0];
__m128i val = _mm_loadu_si128((const __m128i*)(p+1));
- __m128i mask = detail::groupVarintSSEMasks[key];
+ __m128i mask =
+ _mm_load_si128((const __m128i*)&detail::groupVarintSSEMasks[key * 2]);
__m128i r = _mm_shuffle_epi8(val, mask);
// Extracting 32 bits at a time out of an XMM register is a SSE4 feature
-#ifdef __SSE4__
+#if FOLLY_SSE >= 4
*a = _mm_extract_epi32(r, 0);
*b = _mm_extract_epi32(r, 1);
*c = _mm_extract_epi32(r, 2);
} // namespace folly
-#endif /* FOLLY_X64 || defined(__i386__) */
-#endif /* FOLLY_GROUPVARINT_H_ */
+#endif /* FOLLY_X64 || defined(__i386__) || FOLLY_PPC64 */