1 /*
2 * Copyright (c) 2015, 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 * @test
26 * @bug 8073583
27 * @summary C2 support for CRC32C on SPARC
28 *
29 * @run main/othervm/timeout=600 -Xbatch compiler.intrinsics.zip.TestCRC32C -m
30 * @run main/othervm/timeout=600 -Xint -Doffset=1 compiler.intrinsics.zip.TestCRC32C
31 * @run main/othervm/timeout=600 -Xcomp -XX:+TieredCompilation -Doffset=1 compiler.intrinsics.zip.TestCRC32C
32 * @run main/othervm/timeout=600 -Xcomp -XX:-TieredCompilation -Doffset=1 compiler.intrinsics.zip.TestCRC32C
33 */
34
35 package compiler.intrinsics.zip;
36
37 import java.nio.ByteBuffer;
38 import java.util.zip.CRC32C;
39 import java.util.zip.Checksum;
40
41 public class TestCRC32C {
42 // CRC32C (Castagnoli) polynomial
43 // coefficients in different forms
44 // normal: polyBits = 0x1edc6f41 = 0b0001 1110 1101 1100 0110 1111 0100 0001
45 // reversed: polybits = 0x82f63b78 = 0b1000 0010 1111 0110 0011 1011 0111 1000
46 // reversed reciprocal polybits = 0x8f6e37a0 = 0b1000 1111 0110 1110 0011 0111 1010 0000
47 //
48 // 0 5 9 13 17 21 25 29
49 // | | | | | | | |
50 // reversed shiftL 1 polyBits = 0x105ec76f1L = 0b1 0000 0101 1110 1100 0111 0110 1111 0001
51 final static long polyBits = (1L<<(32-32)) + (1L<<(32-28)) + (1L<<(32-27))
52 + (1L<<(32-26)) + (1L<<(32-25)) + (1L<<(32-23)) + (1L<<(32-22))
53 + (1L<<(32-20)) + (1L<<(32-19)) + (1L<<(32-18)) + (1L<<(32-14))
54 + (1L<<(32-13)) + (1L<<(32-11)) + (1L<<(32-10)) + (1L<<(32-9))
55 + (1L<<(32-8)) + (1L<<(32-6)) + (1L<<(32-0));
56 final static long polyBitsShifted = polyBits>>1;
57
58 public static void main(String[] args) {
59 int offset = Integer.getInteger("offset", 0);
60 int msgSize = Integer.getInteger("msgSize", 512);
61 boolean multi = false;
62 int iters = 20000;
63 int warmupIters = 20000;
64
65 if (args.length > 0) {
66 if (args[0].equals("-m")) {
67 multi = true;
68 } else {
69 iters = Integer.valueOf(args[0]);
70 }
71 if (args.length > 1) {
72 warmupIters = Integer.valueOf(args[1]);
73 }
74 }
75
76 if (multi) {
77 test_multi(warmupIters);
78 return;
79 }
80
81 System.out.println(" offset = " + offset);
82 System.out.println("msgSize = " + msgSize + " bytes");
83 System.out.println(" iters = " + iters);
84
85 byte[] b = initializedBytes(msgSize, offset);
86
87 final long crc_reference = update_byteloop(0, b, offset);
88
89 CRC32C crc0 = new CRC32C();
90 CRC32C crc1 = new CRC32C();
91 CRC32C crc2 = new CRC32C();
92
93 crc0.update(b, offset, msgSize);
94 if (!check(crc0, crc_reference)) {
95 System.out.println("CRC32C: crc mismatch during initialization.");
96 return;
97 }
98
99 System.out.println("-------------------------------------------------------");
100
101 /* warm up */
102 for (int i = 0; i < warmupIters; i++) {
103 crc1.reset();
104 crc1.update(b, offset, msgSize);
105 if (!check(crc1, crc_reference)) {
106 System.out.println("CRC32C: crc mismatch during warmup iteration " + i);
107 break;
108 }
109 }
110
111 /* measure performance */
112 long start = System.nanoTime();
113 for (int i = 0; i < iters; i++) {
114 crc1.reset();
115 crc1.update(b, offset, msgSize);
116 }
117 long end = System.nanoTime();
118 double total = (double)(end - start)/1e9; // in seconds
119 double thruput = (double)msgSize*iters/1e6/total; // in MB/s
120 System.out.println("CRC32C.update(byte[]) runtime = " + total + " seconds");
121 System.out.println("CRC32C.update(byte[]) throughput = " + thruput + " MB/s");
122
123 /* check correctness */
124 for (int i = 0; i < iters; i++) {
125 crc1.reset();
126 crc1.update(b, offset, msgSize);
127 if (!check(crc0, crc1)) break;
128 }
129 report("CRCs", crc0, crc1);
130
131 System.out.println("-------------------------------------------------------");
132
133 ByteBuffer buf = ByteBuffer.allocateDirect(msgSize);
134 buf.put(b, offset, msgSize);
135 buf.flip();
136
137 /* warm up */
138 for (int i = 0; i < warmupIters; i++) {
139 crc2.reset();
140 crc2.update(buf);
141 buf.rewind();
142 }
143
144 /* measure performance */
145 start = System.nanoTime();
146 for (int i = 0; i < iters; i++) {
147 crc2.reset();
148 crc2.update(buf);
149 buf.rewind();
150 }
151 end = System.nanoTime();
152 total = (double)(end - start)/1e9; // in seconds
153 thruput = (double)msgSize*iters/1e6/total; // in MB/s
154 System.out.println("CRC32C.update(ByteBuffer) runtime = " + total + " seconds");
155 System.out.println("CRC32C.update(ByteBuffer) throughput = " + thruput + " MB/s");
156
157 /* check correctness */
158 for (int i = 0; i < iters; i++) {
159 crc2.reset();
160 crc2.update(buf);
161 buf.rewind();
162 if (!check(crc0, crc2)) break;
163 }
164 report("CRCs", crc0, crc2);
165
166 System.out.println("-------------------------------------------------------");
167 }
168
169 public static long update_byteloop(long crc, byte[] buf, int offset) {
170 for (int i = offset; i < buf.length; i++) {
171 crc = update_singlebyte(crc, polyBitsShifted, buf[i]);
172 }
173 return crc;
174 }
175
176 // Straight-forward implementation of CRC update by one byte.
177 public static long update_singlebyte(long crc, long polynomial, int val) {
178 crc = (crc ^ -1L) & 0x00000000ffffffffL; // use 1's complement of crc
179 crc = crc ^ (val&0xff); // XOR in next byte from stream
180 for (int i = 0; i < 8; i++) {
181 boolean bitset = (crc & 0x01L) != 0;
182
183 crc = crc>>1;
184 if (bitset) {
185 crc = crc ^ polynomial;
186 crc = crc & 0x00000000ffffffffL;
187 }
188 }
189 crc = (crc ^ -1L) & 0x00000000ffffffffL; // revert taking 1's complement
190 return crc;
191 }
192
193 private static void report(String s, Checksum crc0, Checksum crc1) {
194 System.out.printf("%s: crc0 = %08x, crc1 = %08x\n",
195 s, crc0.getValue(), crc1.getValue());
196 }
197
198 private static boolean check(Checksum crc0, Checksum crc1) {
199 if (crc0.getValue() != crc1.getValue()) {
200 System.err.printf("ERROR: crc0 = %08x, crc1 = %08x\n",
201 crc0.getValue(), crc1.getValue());
202 return false;
203 }
204 return true;
205 }
206
207 private static boolean check(Checksum crc0, long crc_reference) {
208 if (crc0.getValue() != crc_reference) {
209 System.err.printf("ERROR: crc0 = %08x, crc_reference = %08x\n",
210 crc0.getValue(), crc_reference);
211 return false;
212 }
213 return true;
214 }
215
216 private static byte[] initializedBytes(int M, int offset) {
217 byte[] bytes = new byte[M + offset];
218 for (int i = 0; i < offset; i++) {
219 bytes[i] = (byte) i;
220 }
221 for (int i = offset; i < bytes.length; i++) {
222 bytes[i] = (byte) (i - offset);
223 }
224 return bytes;
225 }
226
227 private static void test_multi(int iters) {
228 int len1 = 8; // the 8B/iteration loop
229 int len2 = 32; // the 32B/iteration loop
230 int len3 = 4096; // the 4KB/iteration loop
231
232 byte[] b = initializedBytes(len3*16, 0);
233 int[] offsets = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 16, 32, 64, 128, 256, 512 };
234 int[] sizes = { 0, 1, 2, 3, 4, 5, 6, 7,
235 len1, len1+1, len1+2, len1+3, len1+4, len1+5, len1+6, len1+7,
236 len1*2, len1*2+1, len1*2+3, len1*2+5, len1*2+7,
237 len2, len2+1, len2+3, len2+5, len2+7,
238 len2*2, len2*4, len2*8, len2*16, len2*32, len2*64,
239 len3, len3+1, len3+3, len3+5, len3+7,
240 len3*2, len3*4, len3*8,
241 len1+len2, len1+len2+1, len1+len2+3, len1+len2+5, len1+len2+7,
242 len1+len3, len1+len3+1, len1+len3+3, len1+len3+5, len1+len3+7,
243 len2+len3, len2+len3+1, len2+len3+3, len2+len3+5, len2+len3+7,
244 len1+len2+len3, len1+len2+len3+1, len1+len2+len3+3,
245 len1+len2+len3+5, len1+len2+len3+7,
246 (len1+len2+len3)*2, (len1+len2+len3)*2+1, (len1+len2+len3)*2+3,
247 (len1+len2+len3)*2+5, (len1+len2+len3)*2+7,
248 (len1+len2+len3)*3, (len1+len2+len3)*3-1, (len1+len2+len3)*3-3,
249 (len1+len2+len3)*3-5, (len1+len2+len3)*3-7 };
250 CRC32C[] crc0 = new CRC32C[offsets.length*sizes.length];
251 CRC32C[] crc1 = new CRC32C[offsets.length*sizes.length];
252 int i, j, k;
253
254 System.out.printf("testing %d cases ...\n", offsets.length*sizes.length);
255
256 /* set the result from interpreter as reference */
257 for (i = 0; i < offsets.length; i++) {
258 for (j = 0; j < sizes.length; j++) {
259 crc0[i*sizes.length + j] = new CRC32C();
260 crc1[i*sizes.length + j] = new CRC32C();
261 crc0[i*sizes.length + j].update(b, offsets[i], sizes[j]);
262 }
263 }
264
265 /* warm up the JIT compiler and get result */
266 for (k = 0; k < iters; k++) {
267 for (i = 0; i < offsets.length; i++) {
268 for (j = 0; j < sizes.length; j++) {
269 crc1[i*sizes.length + j].reset();
270 crc1[i*sizes.length + j].update(b, offsets[i], sizes[j]);
271 }
272 }
273 }
274
275 /* check correctness */
276 for (i = 0; i < offsets.length; i++) {
277 for (j = 0; j < sizes.length; j++) {
278 if (!check(crc0[i*sizes.length + j], crc1[i*sizes.length + j])) {
279 System.out.printf("offsets[%d] = %d", i, offsets[i]);
280 System.out.printf("\tsizes[%d] = %d\n", j, sizes[j]);
281 }
282 }
283 }
284 }
285 }