mirror of
https://github.com/golang/go.git
synced 2025-12-08 06:10:04 +00:00
a108b280bc
This change implements the GC pacer redesign outlined in #44167 and the accompanying design document, behind a GOEXPERIMENT flag that is on by default. In addition to adding the new pacer, this CL also includes code to track and account for stack and globals scan work in the pacer and in the assist credit system. The new pacer also deviates slightly from the document in that it increases the bound on the minimum trigger ratio from 0.6 (scaled by GOGC) to 0.7. The logic behind this change is that the new pacer much more consistently hits the goal (good!) leading to slightly less frequent GC cycles, but _longer_ ones (in this case, bad!). It turns out that the cost of having the GC on hurts throughput significantly (per byte of memory used), though tail latencies can improve by up to 10%! To be conservative, this change moves the value to 0.7 where there is a small improvement to both throughput and latency, given the memory use. Because the new pacer accounts for the two most significant sources of scan work after heap objects, it is now also safer to reduce the minimum heap size without leading to very poor amortization. This change thus decreases the minimum heap size to 512 KiB, which corresponds to the fact that the runtime has around 200 KiB of scannable globals always there, up-front, providing a baseline. Benchmark results: https://perf.golang.org/search?q=upload:20211001.6 tile38's KNearest benchmark shows a memory increase, but throughput (and latency) per byte of memory used is better. gopher-lua showed an increase in both CPU time and memory usage, but subsequent attempts to reproduce this behavior are inconsistent. Sometimes the overall performance is better, sometimes it's worse. This suggests that the benchmark is fairly noisy in a way not captured by the benchmarking framework itself. biogo-igor is the only benchmark to show a significant performance loss. This benchmark exhibits a very high GC rate, with relatively little work to do in each cycle. The idle mark workers are quite active. In the new pacer, mark phases are longer, mark assists are fewer, and some of that time in mark assists has shifted to idle workers. Linux perf indicates that the difference in CPU time can be mostly attributed to write-barrier slow path related calls, which in turn indicates that the write barrier being on for longer is the primary culprit. This also explains the memory increase, as a longer mark phase leads to more memory allocated black, surviving an extra cycle and contributing to the heap goal. For #44167. Change-Id: I8ac7cfef7d593e4a642c9b2be43fb3591a8ec9c4 Reviewed-on: https://go-review.googlesource.com/c/go/+/309869 Trust: Michael Knyszek <mknyszek@google.com> Run-TryBot: Michael Knyszek <mknyszek@google.com> TryBot-Result: Go Bot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com> Reviewed-by: Michael Pratt <mpratt@google.com>
92 lines
3.4 KiB
Go
92 lines
3.4 KiB
Go
// Copyright 2021 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 goexperiment implements support for toolchain experiments.
|
|
//
|
|
// Toolchain experiments are controlled by the GOEXPERIMENT
|
|
// environment variable. GOEXPERIMENT is a comma-separated list of
|
|
// experiment names. GOEXPERIMENT can be set at make.bash time, which
|
|
// sets the default experiments for binaries built with the tool
|
|
// chain; or it can be set at build time. GOEXPERIMENT can also be set
|
|
// to "none", which disables any experiments that were enabled at
|
|
// make.bash time.
|
|
//
|
|
// Experiments are exposed to the build in the following ways:
|
|
//
|
|
// - Build tag goexperiment.x is set if experiment x (lower case) is
|
|
// enabled.
|
|
//
|
|
// - For each experiment x (in camel case), this package contains a
|
|
// boolean constant x and an integer constant xInt.
|
|
//
|
|
// - In runtime assembly, the macro GOEXPERIMENT_x is defined if
|
|
// experiment x (lower case) is enabled.
|
|
//
|
|
// In the toolchain, the set of experiments enabled for the current
|
|
// build should be accessed via objabi.Experiment.
|
|
//
|
|
// The set of experiments is included in the output of runtime.Version()
|
|
// and "go version <binary>" if it differs from the default experiments.
|
|
//
|
|
// For the set of experiments supported by the current toolchain, see
|
|
// "go doc goexperiment.Flags".
|
|
//
|
|
// Note that this package defines the set of experiments (in Flags)
|
|
// and records the experiments that were enabled when the package
|
|
// was compiled (as boolean and integer constants).
|
|
//
|
|
// Note especially that this package does not itself change behavior
|
|
// at run time based on the GOEXPERIMENT variable.
|
|
// The code used in builds to interpret the GOEXPERIMENT variable
|
|
// is in the separate package internal/buildcfg.
|
|
package goexperiment
|
|
|
|
//go:generate go run mkconsts.go
|
|
|
|
// Flags is the set of experiments that can be enabled or disabled in
|
|
// the current toolchain.
|
|
//
|
|
// When specified in the GOEXPERIMENT environment variable or as build
|
|
// tags, experiments use the strings.ToLower of their field name.
|
|
//
|
|
// For the baseline experimental configuration, see
|
|
// objabi.experimentBaseline.
|
|
//
|
|
// If you change this struct definition, run "go generate".
|
|
type Flags struct {
|
|
FieldTrack bool
|
|
PreemptibleLoops bool
|
|
StaticLockRanking bool
|
|
|
|
// Unified enables the compiler's unified IR construction
|
|
// experiment.
|
|
Unified bool
|
|
|
|
// Regabi is split into several sub-experiments that can be
|
|
// enabled individually. Not all combinations work.
|
|
// The "regabi" GOEXPERIMENT is an alias for all "working"
|
|
// subexperiments.
|
|
|
|
// RegabiWrappers enables ABI wrappers for calling between
|
|
// ABI0 and ABIInternal functions. Without this, the ABIs are
|
|
// assumed to be identical so cross-ABI calls are direct.
|
|
RegabiWrappers bool
|
|
// RegabiReflect enables the register-passing paths in
|
|
// reflection calls. This is also gated by intArgRegs in
|
|
// reflect and runtime (which are disabled by default) so it
|
|
// can be used in targeted tests.
|
|
RegabiReflect bool
|
|
// RegabiArgs enables register arguments/results in all
|
|
// compiled Go functions.
|
|
//
|
|
// Requires wrappers (to do ABI translation), and reflect (so
|
|
// reflection calls use registers).
|
|
RegabiArgs bool
|
|
|
|
// PacerRedesign enables the new GC pacer in the runtime.
|
|
//
|
|
// Details regarding the new pacer may be found at
|
|
// https://golang.org/design/44167-gc-pacer-redesign
|
|
PacerRedesign bool
|
|
}
|