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
|
/*
* Copyright (c) 2023, 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 8304720
* @summary Test some examples where non-vectorized memops also need to
* be reordered during SuperWord::schedule.
* @modules java.base/jdk.internal.misc
* @library /test/lib /
* @run driver compiler.loopopts.superword.TestScheduleReordersScalarMemops nCOH_nAV
* @run driver compiler.loopopts.superword.TestScheduleReordersScalarMemops nCOH_yAV
* @run driver compiler.loopopts.superword.TestScheduleReordersScalarMemops yCOH_nAV
* @run driver compiler.loopopts.superword.TestScheduleReordersScalarMemops yCOH_yAV
*/
package compiler.loopopts.superword;
import jdk.internal.misc.Unsafe;
import jdk.test.lib.Asserts;
import compiler.lib.ir_framework.*;
public class TestScheduleReordersScalarMemops {
static final int RANGE = 1024;
static final int ITER = 10_000;
static Unsafe unsafe = Unsafe.getUnsafe();
int[] goldI0 = new int[RANGE];
float[] goldF0 = new float[RANGE];
int[] goldI1 = new int[RANGE];
float[] goldF1 = new float[RANGE];
public static void main(String args[]) {
TestFramework framework = new TestFramework(TestScheduleReordersScalarMemops.class);
framework.addFlags("--add-modules", "java.base", "--add-exports", "java.base/jdk.internal.misc=ALL-UNNAMED",
"-XX:CompileCommand=compileonly,compiler.loopopts.superword.TestScheduleReordersScalarMemops::test*",
"-XX:CompileCommand=compileonly,compiler.loopopts.superword.TestScheduleReordersScalarMemops::verify",
"-XX:CompileCommand=compileonly,compiler.loopopts.superword.TestScheduleReordersScalarMemops::init",
"-XX:-TieredCompilation", "-Xbatch",
"-XX:+IgnoreUnrecognizedVMOptions", "-XX:LoopUnrollLimit=1000");
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();
}
TestScheduleReordersScalarMemops() {
// compute the gold standard in interpreter mode
init(goldI0, goldF0);
test0(goldI0, goldI0, goldF0, goldF0);
init(goldI1, goldF1);
test1(goldI1, goldI1, goldF1, goldF1);
}
@Run(test = "test0")
@Warmup(100)
public void runTest0() {
int[] dataI = new int[RANGE];
float[] dataF = new float[RANGE];
init(dataI, dataF);
test0(dataI, dataI, dataF, dataF);
verify("test0", dataI, goldI0);
verify("test0", dataF, goldF0);
}
@Test
@IR(counts = {IRNode.MUL_VI, "> 0"},
applyIfOr = {"UseCompactObjectHeaders", "false", "AlignVector", "false"},
applyIfCPUFeatureOr = {"avx2", "true", "asimd", "true"})
static void test0(int[] dataIa, int[] dataIb, float[] dataFa, float[] dataFb) {
for (int i = 0; i < RANGE; i+=2) {
// We have dependency edges:
// A -> X
// Y -> B
// Still, we can vectorize [X,Y].
// We do not vectorize A and B, because they are not isomorphic (add vs mul).
//
// Imagine this is unrolled at least 2x.
// We get order: A0 X0 Y0 B0 A1 X1 Y1 B1
// Vectorized: X0 Y0 X1 Y1
// Scalar: A0 B0 A1 B1
//
// However, since the As need to be before, and the Bs after the vector operations,
// we need to have all As before all Bs. This means we need to reorder the scalar
// operations, and not just the vectorized ones.
//
// A correct reordering would be: A0 A1 [X0, Y0, X1, Y1] B0 B1
//
dataFa[i + 0] = dataIa[i + 0] * 1.3f; // A *1.3
dataIb[i + 0] = (int)dataFb[i + 0] * 11; // X *11
dataIb[i + 1] = (int)dataFb[i + 1] * 11; // Y *11
dataFa[i + 1] = dataIa[i + 1] + 1.2f; // B +1.2
// With AlignVector, we need 8-byte alignment of vector loads/stores.
// UseCompactObjectHeaders=false UseCompactObjectHeaders=true
// adr = base + 16 + 8*i -> always adr = base + 12 + 8*i -> never
// -> vectorize -> no vectorization
}
}
@Run(test = "test1")
@Warmup(100)
public void runTest1() {
int[] dataI = new int[RANGE];
float[] dataF = new float[RANGE];
init(dataI, dataF);
test1(dataI, dataI, dataF, dataF);
verify("test1", dataI, goldI1);
verify("test1", dataF, goldF1);
}
@Test
@IR(counts = {IRNode.MUL_VI, "> 0"},
applyIfOr = {"UseCompactObjectHeaders", "false", "AlignVector", "false"},
applyIfCPUFeatureOr = {"avx2", "true", "asimd", "true"})
static void test1(int[] dataIa, int[] dataIb, float[] dataFa, float[] dataFb) {
for (int i = 0; i < RANGE; i+=2) {
// Do the same as test0, but without int-float conversion.
// This should reproduce on machines where conversion is not implemented.
unsafe.putInt(dataFa, unsafe.ARRAY_FLOAT_BASE_OFFSET + 4L * i + 0, dataIa[i+0] + 1); // A +1
dataIb[i+0] = 11 * unsafe.getInt(dataFb, unsafe.ARRAY_INT_BASE_OFFSET + 4L * i + 0); // X
dataIb[i+1] = 11 * unsafe.getInt(dataFb, unsafe.ARRAY_INT_BASE_OFFSET + 4L * i + 4); // Y
unsafe.putInt(dataFa, unsafe.ARRAY_FLOAT_BASE_OFFSET + 4L * i + 4, dataIa[i+1] * 11); // B *11
// With AlignVector, we need 8-byte alignment of vector loads/stores.
// UseCompactObjectHeaders=false UseCompactObjectHeaders=true
// adr = base + 16 + 8*i -> always adr = base + 12 + 8*i -> never
// -> vectorize -> no vectorization
}
}
static void init(int[] dataI, float[] dataF) {
for (int i = 0; i < RANGE; i++) {
dataI[i] = i + 1;
dataF[i] = i + 0.1f;
}
}
static void verify(String name, int[] data, int[] gold) {
for (int i = 0; i < RANGE; i++) {
if (data[i] != gold[i]) {
throw new RuntimeException(" Invalid " + name + " result: data[" + i + "]: " + data[i] + " != " + gold[i]);
}
}
}
static void verify(String name, float[] data, float[] gold) {
for (int i = 0; i < RANGE; i++) {
int datav = unsafe.getInt(data, unsafe.ARRAY_FLOAT_BASE_OFFSET + 4 * i);
int goldv = unsafe.getInt(gold, unsafe.ARRAY_FLOAT_BASE_OFFSET + 4 * i);
if (datav != goldv) {
throw new RuntimeException(" Invalid " + name + " result: dataF[" + i + "]: " + datav + " != " + goldv);
}
}
}
}
|
