1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
|
/*
* Copyright (c) 2024, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
/**
* @test
* @bug 8320646
* @summary Auto-vectorize Float.floatToFloat16, Float.float16ToFloat APIs, with NaN
* @requires vm.compiler2.enabled
* @requires (os.arch == "riscv64" & vm.cpu.features ~= ".*zvfh.*")
* @library /test/lib /
* @run driver compiler.vectorization.TestFloatConversionsVectorNaN nCOH_nAV
* @run driver compiler.vectorization.TestFloatConversionsVectorNaN nCOH_yAV
* @run driver compiler.vectorization.TestFloatConversionsVectorNaN yCOH_nAV
* @run driver compiler.vectorization.TestFloatConversionsVectorNaN yCOH_yAV
*/
package compiler.vectorization;
import java.util.HexFormat;
import compiler.lib.ir_framework.*;
import jdk.test.lib.Asserts;
public class TestFloatConversionsVectorNaN {
private static final int ARRLEN = 1024;
private static final int ITERS = 11000;
private static float [] finp;
private static short [] sout;
private static short [] sinp;
private static float [] fout;
public static void main(String args[]) {
TestFramework framework = new TestFramework(TestFloatConversionsVectorNaN.class);
framework.addFlags("-XX:-TieredCompilation", "-XX:CompileThresholdScaling=0.3");
switch (args[0]) {
case "nCOH_nAV" -> { framework.addFlags("-XX:-UseCompactObjectHeaders", "-XX:-AlignVector"); }
case "nCOH_yAV" -> { framework.addFlags("-XX:-UseCompactObjectHeaders", "-XX:+AlignVector"); }
case "yCOH_nAV" -> { framework.addFlags("-XX:+UseCompactObjectHeaders", "-XX:-AlignVector"); }
case "yCOH_yAV" -> { framework.addFlags("-XX:+UseCompactObjectHeaders", "-XX:+AlignVector"); }
default -> { throw new RuntimeException("Test argument not recognized: " + args[0]); }
};
framework.start();
System.out.println("PASSED");
}
@Test
@IR(counts = {IRNode.VECTOR_CAST_F2HF, IRNode.VECTOR_SIZE + "min(max_float, max_short)", "> 0"},
applyIfOr = {"UseCompactObjectHeaders", "false", "AlignVector", "false"})
public void test_float_float16(short[] sout, float[] finp) {
for (int i = 0; i < finp.length; i++) {
sout[i] = Float.floatToFloat16(finp[i]);
// With AlignVector, we need 8-byte alignment of vector loads/stores.
// UseCompactObjectHeaders=false UseCompactObjectHeaders=true
// F_adr = base + 16 + 4*i -> i % 2 = 0 F_adr = base + 12 + 4*i -> i % 2 = 1
// S_adr = base + 16 + 2*i -> i % 4 = 0 S_adr = base + 12 + 2*i -> i % 4 = 2
// -> vectorize -> no vectorization
}
}
@Run(test = {"test_float_float16"}, mode = RunMode.STANDALONE)
public void kernel_test_float_float16() {
int errno = 0;
finp = new float[ARRLEN];
sout = new short[ARRLEN];
// Setup
for (int i = 0; i < ARRLEN; i++) {
if (i%39 == 0) {
int x = 0x7f800000 + ((i/39) << 13);
x = (i%2 == 0) ? x : (x | 0x80000000);
finp[i] = Float.intBitsToFloat(x);
} else {
finp[i] = (float) i * 1.4f;
}
}
int ranges[][] = {
{128, 64},
{256, 19},
{384-19, 19},
{512-19, 17},
{640+19, 19},
{768+19, 32},
{896-19, 32}
};
for (int range[] : ranges) {
int start = range[0];
int offset = range[1];
for (int i = start; i < start+offset; i++) {
int x = 0x7f800000 + (i << 13);
finp[i] = Float.intBitsToFloat(x);
}
}
// Test
for (int i = 0; i < ITERS; i++) {
test_float_float16(sout, finp);
}
// Verifying the result
for (int i = 0; i < ARRLEN; i++) {
errno += assertEquals(i, finp[i], Float.floatToFloat16(finp[i]), sout[i]);
}
if (errno > 0) {
throw new RuntimeException("errors occur");
}
}
static int assertEquals(int idx, float f, short expected, short actual) {
HexFormat hf = HexFormat.of();
String msg = "floatToFloat16 wrong result: idx: " + idx + ", \t" + f +
",\t expected: " + hf.toHexDigits(expected) +
",\t actual: " + hf.toHexDigits(actual);
if ((expected & 0x7c00) != 0x7c00) {
if (expected != actual) {
System.err.println(msg);
return 1;
}
} else if ((expected & 0x3ff) != 0) {
if (((actual & 0x7c00) != 0x7c00) || (actual & 0x3ff) == 0) {
System.err.println(msg);
return 1;
}
}
return 0;
}
@Test
@IR(counts = {IRNode.VECTOR_CAST_HF2F, IRNode.VECTOR_SIZE + "min(max_float, max_short)", "> 0"},
applyIfOr = {"UseCompactObjectHeaders", "false", "AlignVector", "false"})
public void test_float16_float(float[] fout, short[] sinp) {
for (int i = 0; i < sinp.length; i++) {
fout[i] = Float.float16ToFloat(sinp[i]);
// With AlignVector, we need 8-byte alignment of vector loads/stores.
// UseCompactObjectHeaders=false UseCompactObjectHeaders=true
// F_adr = base + 16 + 4*i -> i % 2 = 0 F_adr = base + 12 + 4*i -> i % 2 = 1
// S_adr = base + 16 + 2*i -> i % 4 = 0 S_adr = base + 12 + 2*i -> i % 4 = 2
// -> vectorize -> no vectorization
}
}
@Run(test = {"test_float16_float"}, mode = RunMode.STANDALONE)
public void kernel_test_float16_float() {
int errno = 0;
sinp = new short[ARRLEN];
fout = new float[ARRLEN];
// Setup
for (int i = 0; i < ARRLEN; i++) {
if (i%39 == 0) {
int x = 0x7c00 + i;
x = (i%2 == 0) ? x : (x | 0x8000);
sinp[i] = (short)x;
} else {
sinp[i] = (short)i;
}
}
int ranges[][] = {
{128, 64},
{256, 19},
{384-19, 19},
{512-19, 17},
{640+19, 19},
{768+19, 32},
{896-19, 32}
};
for (int range[] : ranges) {
int start = range[0];
int offset = range[1];
for (int i = start; i < start+offset; i++) {
int x = 0x7c00 + i;
x = (i%2 == 0) ? x : (x | 0x8000);
sinp[i] = (short)x;
}
}
// Test
for (int i = 0; i < ITERS; i++) {
test_float16_float(fout, sinp);
}
// Verifying the result
for (int i = 0; i < ARRLEN; i++) {
errno += assertEquals(i, sinp[i], Float.float16ToFloat(sinp[i]), fout[i]);
}
if (errno > 0) {
throw new RuntimeException("errors occur");
}
}
static int assertEquals(int idx, short s, float expected, float actual) {
String msg = "float16ToFloat wrong result: idx: " + idx + ", \t" + s +
",\t expected: " + expected + ",\t" + Integer.toHexString(Float.floatToIntBits(expected)) +
",\t actual: " + actual + ",\t" + Integer.toHexString(Float.floatToIntBits(actual));
if (!Float.isNaN(expected) || !Float.isNaN(actual)) {
if (expected != actual) {
System.err.println(msg);
return 1;
}
}
return 0;
}
}
|
