ChrisHegarty commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1555874325
Just dumping an initial round of benchmark results, etc, based on what is
currently in this PR.
<details>
<summary>Benchmark source (derived from Robert's)</summary>
```
davekim$ cat src/main/java/testing/DotProductBenchmark.java
package testing;
import org.openjdk.jmh.annotations.*;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.ThreadLocalRandom;
import java.util.concurrent.TimeUnit;
import java.util.stream.IntStream;
import jdk.incubator.vector.FloatVector;
import jdk.incubator.vector.VectorOperators;
import jdk.incubator.vector.VectorSpecies;
@BenchmarkMode(Mode.Throughput)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
@State(Scope.Benchmark)
@Warmup(iterations = 3, time = 3)
@Measurement(iterations = 5, time = 3)
@Fork(value = 1, jvmArgsPrepend = {"--add-modules=jdk.incubator.vector"})
public class DotProductBenchmark {
private float[] a;
private float[] b;
@Param({"1", "4", "6", "8", "13", "16", "25", "32", "64", "100", "128",
"207", "256", "300", "512", "702", "1024"})
//@Param({"1", "4", "6", "8", "13", "16", "25", "32", "64", "100" })
//@Param({"1024"})
int size;
@Setup(Level.Trial)
public void init() {
a = new float[size];
b = new float[size];
for (int i = 0; i < size; ++i) {
a[i] = ThreadLocalRandom.current().nextFloat();
b[i] = ThreadLocalRandom.current().nextFloat();
}
}
static final VectorSpecies<Float> SPECIES = FloatVector.SPECIES_PREFERRED;
@Benchmark
public float dotProductNew() {
if (a.length != b.length) {
throw new IllegalArgumentException("vector dimensions differ: " +
a.length + "!=" + b.length);
}
int i = 0;
float res = 0;
// if the array size is large (2x platform vector size), its worth the
overhead to vectorize
// vector loop is unrolled a single time (2 accumulators in parallel)
if (a.length >= 2 * SPECIES.length()) {
FloatVector acc1 = FloatVector.zero(SPECIES);
FloatVector acc2 = FloatVector.zero(SPECIES);
int upperBound = SPECIES.loopBound(a.length - SPECIES.length());
for (; i < upperBound; i += 2 * SPECIES.length()) {
FloatVector va = FloatVector.fromArray(SPECIES, a, i);
FloatVector vb = FloatVector.fromArray(SPECIES, b, i);
acc1 = acc1.add(va.mul(vb));
FloatVector vc = FloatVector.fromArray(SPECIES, a, i +
SPECIES.length());
FloatVector vd = FloatVector.fromArray(SPECIES, b, i +
SPECIES.length());
acc2 = acc2.add(vc.mul(vd));
}
res += acc1.reduceLanes(VectorOperators.ADD) +
acc2.reduceLanes(VectorOperators.ADD);
}
for (; i < a.length; i++) {
res += b[i] * a[i];
}
return res;
}
@Benchmark
public float dotProductOld() {
if (a.length != b.length) {
throw new IllegalArgumentException("vector dimensions differ: " +
a.length + "!=" + b.length);
}
float res = 0f;
/*
* If length of vector is larger than 8, we use unrolled dot product to
accelerate the
* calculation.
*/
int i;
for (i = 0; i < a.length % 8; i++) {
res += b[i] * a[i];
}
if (a.length < 8) {
return res;
}
for (; i + 31 < a.length; i += 32) {
res +=
b[i + 0] * a[i + 0]
+ b[i + 1] * a[i + 1]
+ b[i + 2] * a[i + 2]
+ b[i + 3] * a[i + 3]
+ b[i + 4] * a[i + 4]
+ b[i + 5] * a[i + 5]
+ b[i + 6] * a[i + 6]
+ b[i + 7] * a[i + 7];
res +=
b[i + 8] * a[i + 8]
+ b[i + 9] * a[i + 9]
+ b[i + 10] * a[i + 10]
+ b[i + 11] * a[i + 11]
+ b[i + 12] * a[i + 12]
+ b[i + 13] * a[i + 13]
+ b[i + 14] * a[i + 14]
+ b[i + 15] * a[i + 15];
res +=
b[i + 16] * a[i + 16]
+ b[i + 17] * a[i + 17]
+ b[i + 18] * a[i + 18]
+ b[i + 19] * a[i + 19]
+ b[i + 20] * a[i + 20]
+ b[i + 21] * a[i + 21]
+ b[i + 22] * a[i + 22]
+ b[i + 23] * a[i + 23];
res +=
b[i + 24] * a[i + 24]
+ b[i + 25] * a[i + 25]
+ b[i + 26] * a[i + 26]
+ b[i + 27] * a[i + 27]
+ b[i + 28] * a[i + 28]
+ b[i + 29] * a[i + 29]
+ b[i + 30] * a[i + 30]
+ b[i + 31] * a[i + 31];
}
for (; i + 7 < a.length; i += 8) {
res +=
b[i + 0] * a[i + 0]
+ b[i + 1] * a[i + 1]
+ b[i + 2] * a[i + 2]
+ b[i + 3] * a[i + 3]
+ b[i + 4] * a[i + 4]
+ b[i + 5] * a[i + 5]
+ b[i + 6] * a[i + 6]
+ b[i + 7] * a[i + 7];
}
return res;
}
}
```
</details>
<details>
<summary>Benchmark results</summary>
```
Benchmark (size) Mode Cnt Score Error
Units
DotProductBenchmark.dotProductNew 1 thrpt 5 486.822 ± 1.260
ops/us
DotProductBenchmark.dotProductOld 1 thrpt 5 547.520 ± 1.362
ops/us
DotProductBenchmark.dotProductNew 4 thrpt 5 276.907 ± 1.612
ops/us
DotProductBenchmark.dotProductOld 4 thrpt 5 398.279 ± 1.195
ops/us
DotProductBenchmark.dotProductNew 6 thrpt 5 273.141 ± 1.060
ops/us
DotProductBenchmark.dotProductOld 6 thrpt 5 364.975 ± 1.939
ops/us
DotProductBenchmark.dotProductNew 8 thrpt 5 219.088 ± 0.538
ops/us
DotProductBenchmark.dotProductOld 8 thrpt 5 273.919 ± 0.897
ops/us
DotProductBenchmark.dotProductNew 13 thrpt 5 186.654 ± 0.075
ops/us
DotProductBenchmark.dotProductOld 13 thrpt 5 199.216 ± 0.476
ops/us
DotProductBenchmark.dotProductNew 16 thrpt 5 160.680 ± 0.401
ops/us
DotProductBenchmark.dotProductOld 16 thrpt 5 155.481 ± 0.382
ops/us
DotProductBenchmark.dotProductNew 25 thrpt 5 103.595 ± 0.358
ops/us
DotProductBenchmark.dotProductOld 25 thrpt 5 99.612 ± 0.262
ops/us
DotProductBenchmark.dotProductNew 32 thrpt 5 84.886 ± 0.342
ops/us
DotProductBenchmark.dotProductOld 32 thrpt 5 103.425 ± 0.364
ops/us
DotProductBenchmark.dotProductNew 64 thrpt 5 78.525 ± 1.889
ops/us
DotProductBenchmark.dotProductOld 64 thrpt 5 53.223 ± 0.108
ops/us
DotProductBenchmark.dotProductNew 100 thrpt 5 60.173 ± 0.453
ops/us
DotProductBenchmark.dotProductOld 100 thrpt 5 32.104 ± 0.027
ops/us
DotProductBenchmark.dotProductNew 128 thrpt 5 64.356 ± 0.145
ops/us
DotProductBenchmark.dotProductOld 128 thrpt 5 27.143 ± 0.083
ops/us
DotProductBenchmark.dotProductNew 207 thrpt 5 35.962 ± 0.015
ops/us
DotProductBenchmark.dotProductOld 207 thrpt 5 16.279 ± 0.253
ops/us
DotProductBenchmark.dotProductNew 256 thrpt 5 49.528 ± 1.180
ops/us
DotProductBenchmark.dotProductOld 256 thrpt 5 13.683 ± 0.137
ops/us
DotProductBenchmark.dotProductNew 300 thrpt 5 36.517 ± 0.104
ops/us
DotProductBenchmark.dotProductOld 300 thrpt 5 11.232 ± 0.007
ops/us
DotProductBenchmark.dotProductNew 512 thrpt 5 36.253 ± 0.004
ops/us
DotProductBenchmark.dotProductOld 512 thrpt 5 6.866 ± 0.078
ops/us
DotProductBenchmark.dotProductNew 702 thrpt 5 17.555 ± 0.184
ops/us
DotProductBenchmark.dotProductOld 702 thrpt 5 4.855 ± 0.020
ops/us
DotProductBenchmark.dotProductNew 1024 thrpt 5 23.363 ± 0.037
ops/us
DotProductBenchmark.dotProductOld 1024 thrpt 5 3.439 ± 0.067
ops/us
```
</details>
<details>
<summary>Machine details (for reference)</summary>
```
davekim$ /home/chegar/binaries/jdk-20.0.1/bin/jshell --add-modules
jdk.incubator.vector
| Welcome to JShell -- Version 20.0.1
| For an introduction type: /help intro
jshell> jdk.incubator.vector.FloatVector.SPECIES_PREFERRED
$1 ==> Species[float, 16, S_512_BIT]
davekim$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=23.04
DISTRIB_CODENAME=lunar
DISTRIB_DESCRIPTION="Ubuntu 23.04"
davekim$ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 39 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 12
On-line CPU(s) list: 0-11
Vendor ID: GenuineIntel
Model name: 11th Gen Intel(R) Core(TM) i5-11400 @ 2.60GHz
CPU family: 6
Model: 167
Thread(s) per core: 2
Core(s) per socket: 6
Socket(s): 1
Stepping: 1
CPU(s) scaling MHz: 62%
CPU max MHz: 4400.0000
CPU min MHz: 800.0000
BogoMIPS: 5184.00
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr
pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall
nx pdpe1gb rdtscp lm constant_tsc
art arch_perfmon pebs bts rep_good nopl xtopology
nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl
vmx est tm2 ssse3 sdbg fma cx16
xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt
tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch
cpuid_fault invpcid_single ssbd ibrs ibpb
stibp ibrs_enhanced tpr_shadow vnmi flexpriority
ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid mpx
avx512f avx512dq rdseed adx smap avx512i
fma clflushopt intel_pt avx512cd sha_ni avx512bw
avx512vl xsaveopt xsavec xgetbv1 xsaves dtherm ida arat pln pts hwp hwp_notify
hwp_act_window hwp_epp hwp_pkg_req
avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes
vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid fsrm md_clear
flush_l1d arch_capabilities
Virtualization features:
Virtualization: VT-x
Caches (sum of all):
L1d: 288 KiB (6 instances)
L1i: 192 KiB (6 instances)
L2: 3 MiB (6 instances)
L3: 12 MiB (1 instance)
NUMA:
NUMA node(s): 1
NUMA node0 CPU(s): 0-11
Vulnerabilities:
Itlb multihit: Not affected
L1tf: Not affected
Mds: Not affected
Meltdown: Not affected
Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Retbleed: Mitigation; Enhanced IBRS
Spec store bypass: Mitigation; Speculative Store Bypass disabled via
prctl
Spectre v1: Mitigation; usercopy/swapgs barriers and __user
pointer sanitization
Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB
filling, PBRSB-eIBRS SW sequence
Srbds: Not affected
Tsx async abort: Not affected
```
</details>
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
