languages/go
languages/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-07-05 09:50:19 +00:00
Code Issues Projects Releases Wiki Activity

runtime: convert flaky semaphore linearity test into benchmark

Browse Source 
Also, add a benchmark for another case that was originally tested.

Also also, remove all the dead code this now creates.

Fixes #53428.

Change-Id: Idbba88d3d31d38a8854fd5ed99001e394da27300
Reviewed-on: https://go-review.googlesource.com/c/go/+/412878
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Michael Knyszek <mknyszek@google.com>
This commit is contained in:
Michael Anthony Knyszek 2022-06-17 19:56:27 +00:00 committed by Gopher Robot
parent 530511bacc
commit 4b236b45d0
6 changed files with 58 additions and 354 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
src/go/build/deps_test.go
View File

@ -543,10 +543,7 @@ var depsRules = `
internal/fuzz, internal/testlog, runtime/pprof, regexp
< testing/internal/testdeps;
MATH, errors, testing
< internal/testmath;
OS, flag, testing, internal/cfg, internal/testmath
OS, flag, testing, internal/cfg
< internal/testenv;
OS, encoding/base64

65
src/internal/testenv/testenv.go
View File

@ -16,7 +16,6 @@ import (
"flag"
"fmt"
"internal/cfg"
"internal/testmath"
"os"
"os/exec"
"path/filepath"
@ -464,67 +463,3 @@ func RunWithTimeout(t testing.TB, cmd *exec.Cmd) ([]byte, error) {
return b.Bytes(), err
}
// CheckLinear checks if the function produced by f scales linearly.
//
// f must accept a scale factor which causes the input to the function it
// produces to scale by that factor.
func CheckLinear(t *testing.T, f func(scale float64) func(*testing.B)) {
MustHaveExec(t)
if os.Getenv("GO_PERF_UNIT_TEST") == "" {
// Invoke the same test as a subprocess with the GO_PERF_UNIT_TEST environment variable set.
// We create a subprocess for two reasons:
//
// 1. There's no other way to set the benchmarking parameters of testing.Benchmark.
// 2. Since we're effectively running a performance test, running in a subprocess grants
// us a little bit more isolation than using the same process.
//
// As an alternative, we could fairly easily reimplement the timing code in testing.Benchmark,
// but a subprocess is just as easy to create.
selfCmd := CleanCmdEnv(exec.Command(os.Args[0], "-test.v", fmt.Sprintf("-test.run=^%s$", t.Name()), "-test.benchtime=1x"))
selfCmd.Env = append(selfCmd.Env, "GO_PERF_UNIT_TEST=1")
output, err := RunWithTimeout(t, selfCmd)
if err != nil {
t.Error(err)
t.Logf("--- subprocess output ---\n%s", string(output))
}
if bytes.Contains(output, []byte("insignificant result")) {
t.Skip("insignificant result")
}
return
}
// Pick a reasonable sample count.
const count = 10
// Collect samples for scale factor 1.
x1 := make([]testing.BenchmarkResult, 0, count)
for i := 0; i < count; i++ {
x1 = append(x1, testing.Benchmark(f(1.0)))
}
// Collect samples for scale factor 2.
x2 := make([]testing.BenchmarkResult, 0, count)
for i := 0; i < count; i++ {
x2 = append(x2, testing.Benchmark(f(2.0)))
}
// Run a t-test on the results.
r1 := testmath.BenchmarkResults(x1)
r2 := testmath.BenchmarkResults(x2)
result, err := testmath.TwoSampleWelchTTest(r1, r2, testmath.LocationDiffers)
if err != nil {
t.Fatalf("failed to run t-test: %v", err)
}
if result.P > 0.005 {
// Insignificant result.
t.Skip("insignificant result")
}
// Let ourselves be within 3x; 2x is too strict.
if m1, m2 := r1.Mean(), r2.Mean(); 3.0*m1 < m2 {
t.Fatalf("failure to scale linearly: µ_1=%s µ_2=%s p=%f", time.Duration(m1), time.Duration(m2), result.P)
}
}

38
src/internal/testmath/bench.go
View File

@ -1,38 +0,0 @@
// Copyright 2022 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package testmath
import (
"math"
"testing"
"time"
)
type BenchmarkResults []testing.BenchmarkResult
func (b BenchmarkResults) Weight() float64 {
var weight int
for _, r := range b {
weight += r.N
}
return float64(weight)
}
func (b BenchmarkResults) Mean() float64 {
var dur time.Duration
for _, r := range b {
dur += r.T * time.Duration(r.N)
}
return float64(dur) / b.Weight()
}
func (b BenchmarkResults) Variance() float64 {
var num float64
mean := b.Mean()
for _, r := range b {
num += math.Pow(float64(r.T)-mean, 2) * float64(r.N)
}
return float64(num) / b.Weight()
}

213
src/internal/testmath/ttest.go
View File

@ -1,213 +0,0 @@
// Copyright 2022 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package testmath
import (
"errors"
"math"
)
// A TTestSample is a sample that can be used for a one or two sample
// t-test.
type TTestSample interface {
Weight() float64
Mean() float64
Variance() float64
}
var (
ErrSampleSize = errors.New("sample is too small")
ErrZeroVariance = errors.New("sample has zero variance")
ErrMismatchedSamples = errors.New("samples have different lengths")
)
// TwoSampleWelchTTest performs a two-sample (unpaired) Welch's t-test
// on samples x1 and x2. This t-test does not assume the distributions
// have equal variance.
func TwoSampleWelchTTest(x1, x2 TTestSample, alt LocationHypothesis) (*TTestResult, error) {
n1, n2 := x1.Weight(), x2.Weight()
if n1 <= 1 || n2 <= 1 {
// TODO: Can we still do this with n == 1?
return nil, ErrSampleSize
}
v1, v2 := x1.Variance(), x2.Variance()
if v1 == 0 && v2 == 0 {
return nil, ErrZeroVariance
}
dof := math.Pow(v1/n1+v2/n2, 2) /
(math.Pow(v1/n1, 2)/(n1-1) + math.Pow(v2/n2, 2)/(n2-1))
s := math.Sqrt(v1/n1 + v2/n2)
t := (x1.Mean() - x2.Mean()) / s
return newTTestResult(int(n1), int(n2), t, dof, alt), nil
}
// A TTestResult is the result of a t-test.
type TTestResult struct {
// N1 and N2 are the sizes of the input samples. For a
// one-sample t-test, N2 is 0.
N1, N2 int
// T is the value of the t-statistic for this t-test.
T float64
// DoF is the degrees of freedom for this t-test.
DoF float64
// AltHypothesis specifies the alternative hypothesis tested
// by this test against the null hypothesis that there is no
// difference in the means of the samples.
AltHypothesis LocationHypothesis
// P is p-value for this t-test for the given null hypothesis.
P float64
}
func newTTestResult(n1, n2 int, t, dof float64, alt LocationHypothesis) *TTestResult {
dist := TDist{dof}
var p float64
switch alt {
case LocationDiffers:
p = 2 * (1 - dist.CDF(math.Abs(t)))
case LocationLess:
p = dist.CDF(t)
case LocationGreater:
p = 1 - dist.CDF(t)
}
return &TTestResult{N1: n1, N2: n2, T: t, DoF: dof, AltHypothesis: alt, P: p}
}
// A LocationHypothesis specifies the alternative hypothesis of a
// location test such as a t-test or a Mann-Whitney U-test. The
// default (zero) value is to test against the alternative hypothesis
// that they differ.
type LocationHypothesis int
const (
// LocationLess specifies the alternative hypothesis that the
// location of the first sample is less than the second. This
// is a one-tailed test.
LocationLess LocationHypothesis = -1
// LocationDiffers specifies the alternative hypothesis that
// the locations of the two samples are not equal. This is a
// two-tailed test.
LocationDiffers LocationHypothesis = 0
// LocationGreater specifies the alternative hypothesis that
// the location of the first sample is greater than the
// second. This is a one-tailed test.
LocationGreater LocationHypothesis = 1
)
// A TDist is a Student's t-distribution with V degrees of freedom.
type TDist struct {
V float64
}
// PDF returns the value at x of the probability distribution function for the
// distribution.
func (t TDist) PDF(x float64) float64 {
return math.Exp(lgamma((t.V+1)/2)-lgamma(t.V/2)) /
math.Sqrt(t.V*math.Pi) * math.Pow(1+(x*x)/t.V, -(t.V+1)/2)
}
// CDF returns the value at x of the cumulative distribution function for the
// distribution.
func (t TDist) CDF(x float64) float64 {
if x == 0 {
return 0.5
} else if x > 0 {
return 1 - 0.5*betaInc(t.V/(t.V+x*x), t.V/2, 0.5)
} else if x < 0 {
return 1 - t.CDF(-x)
} else {
return math.NaN()
}
}
func (t TDist) Bounds() (float64, float64) {
return -4, 4
}
func lgamma(x float64) float64 {
y, _ := math.Lgamma(x)
return y
}
// betaInc returns the value of the regularized incomplete beta
// function Iₓ(a, b) = 1 / B(a, b) * ∫₀ˣ tᵃ⁻¹ (1-t)ᵇ⁻¹ dt.
//
// This is not to be confused with the "incomplete beta function",
// which can be computed as BetaInc(x, a, b)*Beta(a, b).
//
// If x < 0 or x > 1, returns NaN.
func betaInc(x, a, b float64) float64 {
// Based on Numerical Recipes in C, section 6.4. This uses the
// continued fraction definition of I:
//
// (xᵃ*(1-x)ᵇ)/(a*B(a,b)) * (1/(1+(d₁/(1+(d₂/(1+...))))))
//
// where B(a,b) is the beta function and
//
// d_{2m+1} = -(a+m)(a+b+m)x/((a+2m)(a+2m+1))
// d_{2m} = m(b-m)x/((a+2m-1)(a+2m))
if x < 0 || x > 1 {
return math.NaN()
}
bt := 0.0
if 0 < x && x < 1 {
// Compute the coefficient before the continued
// fraction.
bt = math.Exp(lgamma(a+b) - lgamma(a) - lgamma(b) +
a*math.Log(x) + b*math.Log(1-x))
}
if x < (a+1)/(a+b+2) {
// Compute continued fraction directly.
return bt * betacf(x, a, b) / a
} else {
// Compute continued fraction after symmetry transform.
return 1 - bt*betacf(1-x, b, a)/b
}
}
// betacf is the continued fraction component of the regularized
// incomplete beta function Iₓ(a, b).
func betacf(x, a, b float64) float64 {
const maxIterations = 200
const epsilon = 3e-14
raiseZero := func(z float64) float64 {
if math.Abs(z) < math.SmallestNonzeroFloat64 {
return math.SmallestNonzeroFloat64
}
return z
}
c := 1.0
d := 1 / raiseZero(1-(a+b)*x/(a+1))
h := d
for m := 1; m <= maxIterations; m++ {
mf := float64(m)
// Even step of the recurrence.
numer := mf * (b - mf) * x / ((a + 2*mf - 1) * (a + 2*mf))
d = 1 / raiseZero(1+numer*d)
c = raiseZero(1 + numer/c)
h *= d * c
// Odd step of the recurrence.
numer = -(a + mf) * (a + b + mf) * x / ((a + 2*mf) * (a + 2*mf + 1))
d = 1 / raiseZero(1+numer*d)
c = raiseZero(1 + numer/c)
hfac := d * c
h *= hfac
if math.Abs(hfac-1) < epsilon {
return h
}
}
panic("betainc: a or b too big; failed to converge")
}

4
src/runtime/sema.go
View File

@ -35,8 +35,8 @@ import (
// where n is the number of distinct addresses with goroutines blocked
// on them that hash to the given semaRoot.
// See golang.org/issue/17953 for a program that worked badly
// before we introduced the second level of list, and TestSemTableOneAddrCollisionLinear
// for a test that exercises this.
// before we introduced the second level of list, and
// BenchmarkSemTable/OneAddrCollision/* for a benchmark that exercises this.
type semaRoot struct {
lock mutex
treap *sudog // root of balanced tree of unique waiters.

87
src/runtime/sema_test.go
View File

@ -5,7 +5,7 @@
package runtime_test
import (
"internal/testenv"
"fmt"
. "runtime"
"sync"
"sync/atomic"
@ -103,45 +103,68 @@ func testSemaHandoff() bool {
return res == 1 // did the waiter run first?
}
func TestSemTableOneAddrCollisionLinear(t *testing.T) {
testenv.CheckLinear(t, func(scale float64) func(*testing.B) {
n := int(1000 * scale)
return func(b *testing.B) {
func BenchmarkSemTable(b *testing.B) {
for _, n := range []int{1000, 2000, 4000, 8000} {
b.Run(fmt.Sprintf("OneAddrCollision/n=%d", n), func(b *testing.B) {
tab := Escape(new(SemTable))
u := make([]uint32, SemTableSize+1)
b.ResetTimer()
// Simulate two locks colliding on the same semaRoot.
//
// Specifically enqueue all the waiters for the first lock,
// then all the waiters for the second lock.
//
// Then, dequeue all the waiters from the first lock, then
// the second.
//
// Each enqueue/dequeue operation should be O(1), because
// there are exactly 2 locks. This could be O(n) if all
// the waiters for both locks are on the same list, as it
// once was.
for i := 0; i < n; i++ {
if i < n/2 {
tab.Enqueue(&u[0])
} else {
tab.Enqueue(&u[SemTableSize])
for j := 0; j < b.N; j++ {
// Simulate two locks colliding on the same semaRoot.
//
// Specifically enqueue all the waiters for the first lock,
// then all the waiters for the second lock.
//
// Then, dequeue all the waiters from the first lock, then
// the second.
//
// Each enqueue/dequeue operation should be O(1), because
// there are exactly 2 locks. This could be O(n) if all
// the waiters for both locks are on the same list, as it
// once was.
for i := 0; i < n; i++ {
if i < n/2 {
tab.Enqueue(&u[0])
} else {
tab.Enqueue(&u[SemTableSize])
}
}
for i := 0; i < n; i++ {
var ok bool
if i < n/2 {
ok = tab.Dequeue(&u[0])
} else {
ok = tab.Dequeue(&u[SemTableSize])
}
if !ok {
b.Fatal("failed to dequeue")
}
}
}
for i := 0; i < n; i++ {
var ok bool
if i < n/2 {
ok = tab.Dequeue(&u[0])
} else {
ok = tab.Dequeue(&u[SemTableSize])
})
b.Run(fmt.Sprintf("ManyAddrCollision/n=%d", n), func(b *testing.B) {
tab := Escape(new(SemTable))
u := make([]uint32, n*SemTableSize)
b.ResetTimer()
for j := 0; j < b.N; j++ {
// Simulate n locks colliding on the same semaRoot.
//
// Each enqueue/dequeue operation should be O(log n), because
// each semaRoot is a tree. This could be O(n) if it was
// some simpler data structure.
for i := 0; i < n; i++ {
tab.Enqueue(&u[i*SemTableSize])
}
if !ok {
b.Fatal("failed to dequeue")
for i := 0; i < n; i++ {
if !tab.Dequeue(&u[i*SemTableSize]) {
b.Fatal("failed to dequeue")
}
}
}
}
})
})
}
}