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
|
/*
* Copyright (c) 2025, 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 8357530
* @summary Test the effect of AutoVectorizationOverrideProfitability.
* @library /test/lib /
* @run driver compiler.loopopts.superword.TestAutoVectorizationOverrideProfitability
*/
package compiler.loopopts.superword;
import compiler.lib.ir_framework.*;
import compiler.lib.verify.*;
import compiler.lib.generators.Generator;
import static compiler.lib.generators.Generators.G;
public class TestAutoVectorizationOverrideProfitability {
public static final Generator<Integer> GEN_I = G.ints();
public static final Generator<Float> GEN_F = G.floats();
public static int[] aI = new int[10_000];
public static int[] rI = new int[10_000];
public static float[] aF = new float[10_000];
public static float[] rF = new float[10_000];
static {
G.fill(GEN_I, aI);
G.fill(GEN_F, aF);
}
public static void main(String[] args) throws Exception {
// Do not vectorize, even if profitable.
TestFramework.runWithFlags("-XX:+UnlockDiagnosticVMOptions", "-XX:AutoVectorizationOverrideProfitability=0");
// Normal run, i.e. with normal heuristic. In some cases this vectorizes, in some not.
// By default, we have AutoVectorizationOverrideProfitability=1
TestFramework.run();
// Vectorize even if not profitable.
TestFramework.runWithFlags("-XX:+UnlockDiagnosticVMOptions", "-XX:AutoVectorizationOverrideProfitability=2");
}
public static final float GOLD_SIMPLE_FLOAT_REDUCTION = simpleFloatReduction();
@Test
@Warmup(10)
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "= 2"},
counts = {IRNode.ADD_REDUCTION_VF, "> 0"})
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "< 2"},
counts = {IRNode.ADD_REDUCTION_VF, "= 0"})
// The simple float reduction is not profitable. We need to sequentially
// add up the values, and so we cannot move the reduction out of the loop.
private static float simpleFloatReduction() {
float sum = 0;
for (int i = 0; i < aF.length; i++) {
sum += aF[i];
}
return sum;
}
@Check(test="simpleFloatReduction")
public static void checkSimpleFloatReduction(float result) {
Verify.checkEQ(GOLD_SIMPLE_FLOAT_REDUCTION, result);
}
static { simpleFloatCopy(); }
public static final float[] GOLD_SIMPLE_FLOAT_COPY = rF.clone();
@Test
@Warmup(10)
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "> 0"},
counts = {IRNode.LOAD_VECTOR_F, "> 0"})
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "= 0"},
counts = {IRNode.LOAD_VECTOR_F, "= 0"})
// The simple float copy is always profitable.
private static void simpleFloatCopy() {
for (int i = 0; i < aF.length; i++) {
rF[i] = aF[i];
}
}
@Check(test="simpleFloatCopy")
public static void checkSimpleFloatCopy() {
Verify.checkEQ(GOLD_SIMPLE_FLOAT_COPY, rF);
}
public static final int GOLD_SIMPLE_INT_REDUCTION = simpleIntReduction();
@Test
@Warmup(10)
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "= 2"},
counts = {IRNode.ADD_REDUCTION_VI, "> 0", IRNode.ADD_VI, "> 0"})
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "< 2"},
counts = {IRNode.ADD_REDUCTION_VI, "= 0", IRNode.ADD_VI, "= 0"})
// Current heuristics say that this simple int reduction is not profitable.
// But it would actually be profitable, since we are able to move the
// reduction out of the loop (we can reorder the reduction). When moving
// the reduction out of the loop, we instead accumulate with a simple
// ADD_VI inside the loop.
// See: JDK-8307516 JDK-8345044
private static int simpleIntReduction() {
int sum = 0;
for (int i = 0; i < aI.length; i++) {
sum += aI[i];
}
return sum;
}
@Check(test="simpleIntReduction")
public static void checkSimpleIntReduction(int result) {
Verify.checkEQ(GOLD_SIMPLE_INT_REDUCTION, result);
}
static { simpleIntCopy(); }
public static final int[] GOLD_SIMPLE_INT_COPY = rI.clone();
@Test
@Warmup(10)
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "> 0"},
counts = {IRNode.LOAD_VECTOR_I, "> 0"})
@IR(applyIfCPUFeatureOr = {"avx", "true"},
applyIf = {"AutoVectorizationOverrideProfitability", "= 0"},
counts = {IRNode.LOAD_VECTOR_I, "= 0"})
// The simple int copy is always profitable.
private static void simpleIntCopy() {
for (int i = 0; i < aI.length; i++) {
rI[i] = aI[i];
}
}
@Check(test="simpleIntCopy")
public static void checkSimpleIntCopy() {
Verify.checkEQ(GOLD_SIMPLE_INT_COPY, rI);
}
}
|
