diff options
Diffstat (limited to 'absl/strings/internal/fastmem.h')
-rw-r--r-- | absl/strings/internal/fastmem.h | 215 |
1 files changed, 215 insertions, 0 deletions
diff --git a/absl/strings/internal/fastmem.h b/absl/strings/internal/fastmem.h new file mode 100644 index 000000000000..9989b12e34d3 --- /dev/null +++ b/absl/strings/internal/fastmem.h @@ -0,0 +1,215 @@ +// Copyright 2017 The Abseil Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +// +// Fast memory copying and comparison routines. +// strings::fastmemcmp_inlined() replaces memcmp() +// strings::memcpy_inlined() replaces memcpy() +// strings::memeq(a, b, n) replaces memcmp(a, b, n) == 0 +// +// strings::*_inlined() routines are inline versions of the +// routines exported by this module. Sometimes using the inlined +// versions is faster. Measure before using the inlined versions. +// + +#ifndef ABSL_STRINGS_INTERNAL_FASTMEM_H_ +#define ABSL_STRINGS_INTERNAL_FASTMEM_H_ + +#ifdef __SSE4_1__ +#include <immintrin.h> +#endif +#include <cstddef> +#include <cstdint> +#include <cstdio> +#include <cstring> + +#include "absl/base/internal/unaligned_access.h" +#include "absl/base/macros.h" +#include "absl/base/port.h" + +namespace absl { +namespace strings_internal { + +// Return true if the n bytes at a equal the n bytes at b. +// The regions are allowed to overlap. +// +// The performance is similar to the performance of memcmp(), but faster for +// moderately-sized inputs, or inputs that share a common prefix and differ +// somewhere in their last 8 bytes. Further optimizations can be added later +// if it makes sense to do so. Alternatively, if the compiler & runtime improve +// to eliminate the need for this, we can remove it. +inline bool memeq(const char* a, const char* b, size_t n) { + size_t n_rounded_down = n & ~static_cast<size_t>(7); + if (ABSL_PREDICT_FALSE(n_rounded_down == 0)) { // n <= 7 + return memcmp(a, b, n) == 0; + } + // n >= 8 + { + uint64_t u = + ABSL_INTERNAL_UNALIGNED_LOAD64(a) ^ ABSL_INTERNAL_UNALIGNED_LOAD64(b); + uint64_t v = ABSL_INTERNAL_UNALIGNED_LOAD64(a + n - 8) ^ + ABSL_INTERNAL_UNALIGNED_LOAD64(b + n - 8); + if ((u | v) != 0) { // The first or last 8 bytes differ. + return false; + } + } + // The next line forces n to be a multiple of 8. + n = n_rounded_down; + if (n >= 80) { + // In 2013 or later, this should be fast on long strings. + return memcmp(a, b, n) == 0; + } + // Now force n to be a multiple of 16. Arguably, a "switch" would be smart + // here, but there's a difficult-to-evaluate code size vs. speed issue. The + // current approach often re-compares some bytes (worst case is if n initially + // was 16, 32, 48, or 64), but is fairly short. + size_t e = n & 8; + a += e; + b += e; + n -= e; + // n is now in {0, 16, 32, ...}. Process 0 or more 16-byte chunks. + while (n > 0) { +#ifdef __SSE4_1__ + __m128i u = + _mm_xor_si128(_mm_loadu_si128(reinterpret_cast<const __m128i*>(a)), + _mm_loadu_si128(reinterpret_cast<const __m128i*>(b))); + if (!_mm_test_all_zeros(u, u)) { + return false; + } +#else + uint64_t x = + ABSL_INTERNAL_UNALIGNED_LOAD64(a) ^ ABSL_INTERNAL_UNALIGNED_LOAD64(b); + uint64_t y = ABSL_INTERNAL_UNALIGNED_LOAD64(a + 8) ^ + ABSL_INTERNAL_UNALIGNED_LOAD64(b + 8); + if ((x | y) != 0) { + return false; + } +#endif + a += 16; + b += 16; + n -= 16; + } + return true; +} + +inline int fastmemcmp_inlined(const void* va, const void* vb, size_t n) { + const unsigned char* pa = static_cast<const unsigned char*>(va); + const unsigned char* pb = static_cast<const unsigned char*>(vb); + switch (n) { + default: + return memcmp(va, vb, n); + case 7: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 6: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 5: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 4: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 3: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 2: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ++pa; + ++pb; + ABSL_FALLTHROUGH_INTENDED; + case 1: + if (*pa != *pb) return *pa < *pb ? -1 : +1; + ABSL_FALLTHROUGH_INTENDED; + case 0: + break; + } + return 0; +} + +// The standard memcpy operation is slow for variable small sizes. +// This implementation inlines the optimal realization for sizes 1 to 16. +// To avoid code bloat don't use it in case of not performance-critical spots, +// nor when you don't expect very frequent values of size <= 16. +inline void memcpy_inlined(char* dst, const char* src, size_t size) { + // Compiler inlines code with minimal amount of data movement when third + // parameter of memcpy is a constant. + switch (size) { + case 1: + memcpy(dst, src, 1); + break; + case 2: + memcpy(dst, src, 2); + break; + case 3: + memcpy(dst, src, 3); + break; + case 4: + memcpy(dst, src, 4); + break; + case 5: + memcpy(dst, src, 5); + break; + case 6: + memcpy(dst, src, 6); + break; + case 7: + memcpy(dst, src, 7); + break; + case 8: + memcpy(dst, src, 8); + break; + case 9: + memcpy(dst, src, 9); + break; + case 10: + memcpy(dst, src, 10); + break; + case 11: + memcpy(dst, src, 11); + break; + case 12: + memcpy(dst, src, 12); + break; + case 13: + memcpy(dst, src, 13); + break; + case 14: + memcpy(dst, src, 14); + break; + case 15: + memcpy(dst, src, 15); + break; + case 16: + memcpy(dst, src, 16); + break; + default: + memcpy(dst, src, size); + break; + } +} + +} // namespace strings_internal +} // namespace absl + +#endif // ABSL_STRINGS_INTERNAL_FASTMEM_H_ |