diff options
Diffstat (limited to 'src')
-rw-r--r-- | src/vppinfra.am | 3 | ||||
-rw-r--r-- | src/vppinfra/memcpy_avx2.h (renamed from src/vppinfra/memcpy_avx.h) | 141 | ||||
-rw-r--r-- | src/vppinfra/memcpy_avx512.h | 277 | ||||
-rw-r--r-- | src/vppinfra/memcpy_sse3.h | 62 | ||||
-rw-r--r-- | src/vppinfra/string.h | 6 |
5 files changed, 359 insertions, 130 deletions
diff --git a/src/vppinfra.am b/src/vppinfra.am index 8fcae86f88e..cb8748ca9c9 100644 --- a/src/vppinfra.am +++ b/src/vppinfra.am @@ -219,7 +219,8 @@ nobase_include_HEADERS = \ vppinfra/md5.h \ vppinfra/mem.h \ vppinfra/memcpy_sse3.h \ - vppinfra/memcpy_avx.h \ + vppinfra/memcpy_avx2.h \ + vppinfra/memcpy_avx512.h \ vppinfra/mhash.h \ vppinfra/mheap.h \ vppinfra/mheap_bootstrap.h \ diff --git a/src/vppinfra/memcpy_avx.h b/src/vppinfra/memcpy_avx2.h index e987d044b58..da398a4512a 100644 --- a/src/vppinfra/memcpy_avx.h +++ b/src/vppinfra/memcpy_avx2.h @@ -45,8 +45,8 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -#ifndef included_clib_memcpy_avx_h -#define included_clib_memcpy_avx_h +#ifndef included_clib_memcpy_avx2_h +#define included_clib_memcpy_avx2_h #include <stdint.h> #include <x86intrin.h> @@ -84,65 +84,27 @@ clib_mov128 (u8 * dst, const u8 * src) } static inline void -clib_mov256 (u8 * dst, const u8 * src) +clib_mov128blocks (u8 * dst, const u8 * src, size_t n) { - clib_mov128 ((u8 *) dst + 0 * 128, (const u8 *) src + 0 * 128); - clib_mov128 ((u8 *) dst + 1 * 128, (const u8 *) src + 1 * 128); -} - -static inline void -clib_mov64blocks (u8 * dst, const u8 * src, size_t n) -{ - __m256i ymm0, ymm1; + __m256i ymm0, ymm1, ymm2, ymm3; - while (n >= 64) + while (n >= 128) { ymm0 = _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 0 * 32)); - n -= 64; - ymm1 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 1 * 32)); - src = (const u8 *) src + 64; - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 0 * 32), ymm0); - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 1 * 32), ymm1); - dst = (u8 *) dst + 64; - } -} - -static inline void -clib_mov256blocks (u8 * dst, const u8 * src, size_t n) -{ - __m256i ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7; - - while (n >= 256) - { - ymm0 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 0 * 32)); - n -= 256; + n -= 128; ymm1 = _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 1 * 32)); ymm2 = _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 2 * 32)); ymm3 = _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 3 * 32)); - ymm4 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 4 * 32)); - ymm5 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 5 * 32)); - ymm6 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 6 * 32)); - ymm7 = - _mm256_loadu_si256 ((const __m256i *) ((const u8 *) src + 7 * 32)); - src = (const u8 *) src + 256; + src = (const u8 *) src + 128; _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 0 * 32), ymm0); _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 1 * 32), ymm1); _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 2 * 32), ymm2); _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 3 * 32), ymm3); - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 4 * 32), ymm4); - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 5 * 32), ymm5); - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 6 * 32), ymm6); - _mm256_storeu_si256 ((__m256i *) ((u8 *) dst + 7 * 32), ymm7); - dst = (u8 *) dst + 256; + dst = (u8 *) dst + 128; } } @@ -155,9 +117,9 @@ clib_memcpy (void *dst, const void *src, size_t n) size_t dstofss; size_t bits; - /** - * Copy less than 16 bytes - */ + /** + * Copy less than 16 bytes + */ if (n < 16) { if (n & 0x01) @@ -168,47 +130,47 @@ clib_memcpy (void *dst, const void *src, size_t n) } if (n & 0x02) { - *(uint16_t *) dstu = *(const uint16_t *) srcu; - srcu = (uword) ((const uint16_t *) srcu + 1); - dstu = (uword) ((uint16_t *) dstu + 1); + *(u16 *) dstu = *(const u16 *) srcu; + srcu = (uword) ((const u16 *) srcu + 1); + dstu = (uword) ((u16 *) dstu + 1); } if (n & 0x04) { - *(uint32_t *) dstu = *(const uint32_t *) srcu; - srcu = (uword) ((const uint32_t *) srcu + 1); - dstu = (uword) ((uint32_t *) dstu + 1); + *(u32 *) dstu = *(const u32 *) srcu; + srcu = (uword) ((const u32 *) srcu + 1); + dstu = (uword) ((u32 *) dstu + 1); } if (n & 0x08) { - *(uint64_t *) dstu = *(const uint64_t *) srcu; + *(u64 *) dstu = *(const u64 *) srcu; } return ret; } - /** - * Fast way when copy size doesn't exceed 512 bytes - */ + /** + * Fast way when copy size doesn't exceed 512 bytes + */ if (n <= 32) { clib_mov16 ((u8 *) dst, (const u8 *) src); clib_mov16 ((u8 *) dst - 16 + n, (const u8 *) src - 16 + n); return ret; } + if (n <= 48) + { + clib_mov16 ((u8 *) dst, (const u8 *) src); + clib_mov16 ((u8 *) dst + 16, (const u8 *) src + 16); + clib_mov16 ((u8 *) dst - 16 + n, (const u8 *) src - 16 + n); + return ret; + } if (n <= 64) { clib_mov32 ((u8 *) dst, (const u8 *) src); clib_mov32 ((u8 *) dst - 32 + n, (const u8 *) src - 32 + n); return ret; } - if (n <= 512) + if (n <= 256) { - if (n >= 256) - { - n -= 256; - clib_mov256 ((u8 *) dst, (const u8 *) src); - src = (const u8 *) src + 256; - dst = (u8 *) dst + 256; - } if (n >= 128) { n -= 128; @@ -216,6 +178,7 @@ clib_memcpy (void *dst, const void *src, size_t n) src = (const u8 *) src + 128; dst = (u8 *) dst + 128; } + COPY_BLOCK_128_BACK31: if (n >= 64) { n -= 64; @@ -223,7 +186,6 @@ clib_memcpy (void *dst, const void *src, size_t n) src = (const u8 *) src + 64; dst = (u8 *) dst + 64; } - COPY_BLOCK_64_BACK31: if (n > 32) { clib_mov32 ((u8 *) dst, (const u8 *) src); @@ -237,9 +199,9 @@ clib_memcpy (void *dst, const void *src, size_t n) return ret; } - /** - * Make store aligned when copy size exceeds 512 bytes - */ + /** + * Make store aligned when copy size exceeds 256 bytes + */ dstofss = (uword) dst & 0x1F; if (dstofss > 0) { @@ -250,41 +212,24 @@ clib_memcpy (void *dst, const void *src, size_t n) dst = (u8 *) dst + dstofss; } - /** - * Copy 256-byte blocks. - * Use copy block function for better instruction order control, - * which is important when load is unaligned. - */ - clib_mov256blocks ((u8 *) dst, (const u8 *) src, n); + /** + * Copy 128-byte blocks. + */ + clib_mov128blocks ((u8 *) dst, (const u8 *) src, n); bits = n; - n = n & 255; + n = n & 127; bits -= n; src = (const u8 *) src + bits; dst = (u8 *) dst + bits; - /** - * Copy 64-byte blocks. - * Use copy block function for better instruction order control, - * which is important when load is unaligned. - */ - if (n >= 64) - { - clib_mov64blocks ((u8 *) dst, (const u8 *) src, n); - bits = n; - n = n & 63; - bits -= n; - src = (const u8 *) src + bits; - dst = (u8 *) dst + bits; - } - - /** - * Copy whatever left - */ - goto COPY_BLOCK_64_BACK31; + /** + * Copy whatever left + */ + goto COPY_BLOCK_128_BACK31; } -#endif /* included_clib_mamcpy_avx_h */ +#endif /* included_clib_mamcpy_avx2_h */ /* diff --git a/src/vppinfra/memcpy_avx512.h b/src/vppinfra/memcpy_avx512.h new file mode 100644 index 00000000000..003908026c3 --- /dev/null +++ b/src/vppinfra/memcpy_avx512.h @@ -0,0 +1,277 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * 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. + */ +/*- + * BSD LICENSE + * + * Copyright(c) 2010-2014 Intel Corporation. All rights reserved. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the + * distribution. + * * Neither the name of Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef included_clib_memcpy_avx512_h +#define included_clib_memcpy_avx512_h + +#include <stdint.h> +#include <x86intrin.h> + +static inline void +clib_mov16 (u8 * dst, const u8 * src) +{ + __m128i xmm0; + + xmm0 = _mm_loadu_si128 ((const __m128i *) src); + _mm_storeu_si128 ((__m128i *) dst, xmm0); +} + +static inline void +clib_mov32 (u8 * dst, const u8 * src) +{ + __m256i ymm0; + + ymm0 = _mm256_loadu_si256 ((const __m256i *) src); + _mm256_storeu_si256 ((__m256i *) dst, ymm0); +} + +static inline void +clib_mov64 (u8 * dst, const u8 * src) +{ + __m512i zmm0; + + zmm0 = _mm512_loadu_si512 ((const void *) src); + _mm512_storeu_si512 ((void *) dst, zmm0); +} + +static inline void +clib_mov128 (u8 * dst, const u8 * src) +{ + clib_mov64 (dst + 0 * 64, src + 0 * 64); + clib_mov64 (dst + 1 * 64, src + 1 * 64); +} + +static inline void +clib_mov256 (u8 * dst, const u8 * src) +{ + clib_mov128 (dst + 0 * 128, src + 0 * 128); + clib_mov128 (dst + 1 * 128, src + 1 * 128); +} + +static inline void +clib_mov128blocks (u8 * dst, const u8 * src, size_t n) +{ + __m512i zmm0, zmm1; + + while (n >= 128) + { + zmm0 = _mm512_loadu_si512 ((const void *) (src + 0 * 64)); + n -= 128; + zmm1 = _mm512_loadu_si512 ((const void *) (src + 1 * 64)); + src = src + 128; + _mm512_storeu_si512 ((void *) (dst + 0 * 64), zmm0); + _mm512_storeu_si512 ((void *) (dst + 1 * 64), zmm1); + dst = dst + 128; + } +} + +static inline void +clib_mov512blocks (u8 * dst, const u8 * src, size_t n) +{ + __m512i zmm0, zmm1, zmm2, zmm3, zmm4, zmm5, zmm6, zmm7; + + while (n >= 512) + { + zmm0 = _mm512_loadu_si512 ((const void *) (src + 0 * 64)); + n -= 512; + zmm1 = _mm512_loadu_si512 ((const void *) (src + 1 * 64)); + zmm2 = _mm512_loadu_si512 ((const void *) (src + 2 * 64)); + zmm3 = _mm512_loadu_si512 ((const void *) (src + 3 * 64)); + zmm4 = _mm512_loadu_si512 ((const void *) (src + 4 * 64)); + zmm5 = _mm512_loadu_si512 ((const void *) (src + 5 * 64)); + zmm6 = _mm512_loadu_si512 ((const void *) (src + 6 * 64)); + zmm7 = _mm512_loadu_si512 ((const void *) (src + 7 * 64)); + src = src + 512; + _mm512_storeu_si512 ((void *) (dst + 0 * 64), zmm0); + _mm512_storeu_si512 ((void *) (dst + 1 * 64), zmm1); + _mm512_storeu_si512 ((void *) (dst + 2 * 64), zmm2); + _mm512_storeu_si512 ((void *) (dst + 3 * 64), zmm3); + _mm512_storeu_si512 ((void *) (dst + 4 * 64), zmm4); + _mm512_storeu_si512 ((void *) (dst + 5 * 64), zmm5); + _mm512_storeu_si512 ((void *) (dst + 6 * 64), zmm6); + _mm512_storeu_si512 ((void *) (dst + 7 * 64), zmm7); + dst = dst + 512; + } +} + +static inline void * +clib_memcpy (void *dst, const void *src, size_t n) +{ + uword dstu = (uword) dst; + uword srcu = (uword) src; + void *ret = dst; + size_t dstofss; + size_t bits; + + /** + * Copy less than 16 bytes + */ + if (n < 16) + { + if (n & 0x01) + { + *(u8 *) dstu = *(const u8 *) srcu; + srcu = (uword) ((const u8 *) srcu + 1); + dstu = (uword) ((u8 *) dstu + 1); + } + if (n & 0x02) + { + *(u16 *) dstu = *(const u16 *) srcu; + srcu = (uword) ((const u16 *) srcu + 1); + dstu = (uword) ((u16 *) dstu + 1); + } + if (n & 0x04) + { + *(u32 *) dstu = *(const u32 *) srcu; + srcu = (uword) ((const u32 *) srcu + 1); + dstu = (uword) ((u32 *) dstu + 1); + } + if (n & 0x08) + *(u64 *) dstu = *(const u64 *) srcu; + return ret; + } + + /** + * Fast way when copy size doesn't exceed 512 bytes + */ + if (n <= 32) + { + clib_mov16 ((u8 *) dst, (const u8 *) src); + clib_mov16 ((u8 *) dst - 16 + n, (const u8 *) src - 16 + n); + return ret; + } + if (n <= 64) + { + clib_mov32 ((u8 *) dst, (const u8 *) src); + clib_mov32 ((u8 *) dst - 32 + n, (const u8 *) src - 32 + n); + return ret; + } + if (n <= 512) + { + if (n >= 256) + { + n -= 256; + clib_mov256 ((u8 *) dst, (const u8 *) src); + src = (const u8 *) src + 256; + dst = (u8 *) dst + 256; + } + if (n >= 128) + { + n -= 128; + clib_mov128 ((u8 *) dst, (const u8 *) src); + src = (const u8 *) src + 128; + dst = (u8 *) dst + 128; + } + COPY_BLOCK_128_BACK63: + if (n > 64) + { + clib_mov64 ((u8 *) dst, (const u8 *) src); + clib_mov64 ((u8 *) dst - 64 + n, (const u8 *) src - 64 + n); + return ret; + } + if (n > 0) + clib_mov64 ((u8 *) dst - 64 + n, (const u8 *) src - 64 + n); + return ret; + } + + /** + * Make store aligned when copy size exceeds 512 bytes + */ + dstofss = (uword) dst & 0x3F; + if (dstofss > 0) + { + dstofss = 64 - dstofss; + n -= dstofss; + clib_mov64 ((u8 *) dst, (const u8 *) src); + src = (const u8 *) src + dstofss; + dst = (u8 *) dst + dstofss; + } + + /** + * Copy 512-byte blocks. + * Use copy block function for better instruction order control, + * which is important when load is unaligned. + */ + clib_mov512blocks ((u8 *) dst, (const u8 *) src, n); + bits = n; + n = n & 511; + bits -= n; + src = (const u8 *) src + bits; + dst = (u8 *) dst + bits; + + /** + * Copy 128-byte blocks. + * Use copy block function for better instruction order control, + * which is important when load is unaligned. + */ + if (n >= 128) + { + clib_mov128blocks ((u8 *) dst, (const u8 *) src, n); + bits = n; + n = n & 127; + bits -= n; + src = (const u8 *) src + bits; + dst = (u8 *) dst + bits; + } + + /** + * Copy whatever left + */ + goto COPY_BLOCK_128_BACK63; +} + + +#endif /* included_clib_mamcpy_avx512_h */ + + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ diff --git a/src/vppinfra/memcpy_sse3.h b/src/vppinfra/memcpy_sse3.h index f61396c8922..5e4bf7d79c5 100644 --- a/src/vppinfra/memcpy_sse3.h +++ b/src/vppinfra/memcpy_sse3.h @@ -222,9 +222,9 @@ clib_memcpy (void *dst, const void *src, size_t n) return ret; } - /** - * Fast way when copy size doesn't exceed 512 bytes - */ + /** + * Fast way when copy size doesn't exceed 512 bytes + */ if (n <= 32) { clib_mov16 ((u8 *) dst, (const u8 *) src); @@ -295,27 +295,31 @@ clib_memcpy (void *dst, const void *src, size_t n) return ret; } - /** - * Make store aligned when copy size exceeds 512 bytes, - * and make sure the first 15 bytes are copied, because - * unaligned copy functions require up to 15 bytes - * backwards access. - */ - dstofss = 16 - ((uword) dst & 0x0F) + 16; - n -= dstofss; - clib_mov32 ((u8 *) dst, (const u8 *) src); - src = (const u8 *) src + dstofss; - dst = (u8 *) dst + dstofss; + /** + * Make store aligned when copy size exceeds 512 bytes, + * and make sure the first 15 bytes are copied, because + * unaligned copy functions require up to 15 bytes + * backwards access. + */ + dstofss = (uword) dst & 0x0F; + if (dstofss > 0) + { + dstofss = 16 - dstofss + 16; + n -= dstofss; + clib_mov32 ((u8 *) dst, (const u8 *) src); + src = (const u8 *) src + dstofss; + dst = (u8 *) dst + dstofss; + } srcofs = ((uword) src & 0x0F); - /** - * For aligned copy - */ + /** + * For aligned copy + */ if (srcofs == 0) { - /** - * Copy 256-byte blocks - */ + /** + * Copy 256-byte blocks + */ for (; n >= 256; n -= 256) { clib_mov256 ((u8 *) dst, (const u8 *) src); @@ -323,20 +327,20 @@ clib_memcpy (void *dst, const void *src, size_t n) src = (const u8 *) src + 256; } - /** - * Copy whatever left - */ + /** + * Copy whatever left + */ goto COPY_BLOCK_255_BACK15; } - /** - * For copy with unaligned load - */ + /** + * For copy with unaligned load + */ CLIB_MVUNALIGN_LEFT47 (dst, src, n, srcofs); - /** - * Copy whatever left - */ + /** + * Copy whatever left + */ goto COPY_BLOCK_64_BACK15; } diff --git a/src/vppinfra/string.h b/src/vppinfra/string.h index 0e0627d088d..7873e4647d9 100644 --- a/src/vppinfra/string.h +++ b/src/vppinfra/string.h @@ -65,8 +65,10 @@ void clib_memswap (void *_a, void *_b, uword bytes); * so don't let it anywhere near them. */ #ifndef __COVERITY__ -#if __AVX__ -#include <vppinfra/memcpy_avx.h> +#if __AVX512F__ +#include <vppinfra/memcpy_avx512.h> +#elif __AVX2__ +#include <vppinfra/memcpy_avx2.h> #elif __SSSE3__ #include <vppinfra/memcpy_sse3.h> #else |