1 /* 2 * Copyright (c) 2015, 2019, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #ifndef CPU_X86_CRC32C_H 26 #define CPU_X86_CRC32C_H 27 28 enum { 29 // S. Gueron / Information Processing Letters 112 (2012) 184 30 // shows than anything above 6K and below 32K is a good choice 31 // 32K does not deliver any further performance gains 32 // 6K=8*256 (*3 as we compute 3 blocks together) 33 // 34 // Thus selecting the smallest value so it could apply to the largest number 35 // of buffer sizes. 36 CRC32C_HIGH = 8 * 256, 37 38 // empirical 39 // based on ubench study using methodology described in 40 // V. Gopal et al. / Fast CRC Computation for iSCSI Polynomial Using CRC32 Instruction April 2011 8 41 // 42 // arbitrary value between 27 and 256 43 CRC32C_MIDDLE = 8 * 86, 44 45 // V. Gopal et al. / Fast CRC Computation for iSCSI Polynomial Using CRC32 Instruction April 2011 9 46 // shows that 240 and 1024 are equally good choices as the 216==8*27 47 // 48 // Selecting the smallest value which resulted in a significant performance improvement over 49 // sequential version 50 CRC32C_LOW = 8 * 27, 51 52 CRC32C_NUM_ChunkSizeInBytes = 3, 53 54 // We need to compute powers of 64N and 128N for each "chunk" size 55 CRC32C_NUM_PRECOMPUTED_CONSTANTS = ( 2 * CRC32C_NUM_ChunkSizeInBytes ) 56 }; 57 // Notes: 58 // 1. Why we need to choose a "chunk" approach? 59 // Overhead of computing a powers and powers of for an arbitrary buffer of size N is significant 60 // (implementation approaches a library perf.) 61 // 2. Why only 3 "chunks"? 62 // Performance experiments results showed that a HIGH+LOW was not delivering a stable speedup 63 // curve. 64 // 65 // Disclaimer: 66 // If you ever decide to increase/decrease number of "chunks" be sure to modify 67 // a) constants table generation (hotspot/src/cpu/x86/vm/stubRoutines_x86.cpp) 68 // b) constant fetch from that table (macroAssembler_x86.cpp) 69 // c) unrolled for loop (macroAssembler_x86.cpp) 70 71 #endif /* !CPU_X86_CRC32C_H */ 72